docs(discovery): publish provisional feature baseline

* fix(rag): skip hidden and junk directories when indexing (#5559)

index_personal_documents walked the whole tree with no pruning, so
pointing RAG at a real-world folder silently swept in .obsidian/ plugin
JS, .git/ internals, node_modules/, and __pycache__/ — multiplying
indexing time and polluting retrieval with junk chunks.

Prune hidden directories and well-known junk directories from the walk,
and skip hidden files. The explicitly passed root is exempt, so a user
who deliberately indexes a hidden directory still gets its contents.

* fix(rag): prune hidden/junk dirs in the keyword index too, via a shared helper

The #5559 fix pruned only VectorRAG.index_personal_documents (the vector index).
The parallel keyword index built by PersonalDocsManager.refresh_index ->
load_personal_index walked the same tree unpruned, so .obsidian/, .git/,
node_modules/ etc. still swept into keyword retrieval and the file listing —
the 'end-to-end' guarantee was only half true.

Single-source the pruning policy in src/index_walk (prune_index_dirs +
is_indexable_file) and use it from both walkers so they cannot drift again.
The junk-dir match is now case-insensitive, so a Node_Modules on a
case-insensitive filesystem is pruned too.

Tests: keyword-path regressions covering hidden/junk dirs, hidden files, junk
at depth (not just top level), case-insensitive junk, and the explicit-hidden-
root exemption. The existing vector tests still pass against the shared helper.

docs/discovery/BASELINE-STATUS.md

@@ -0,0 +1,62 @@
+# Discovery Baseline Status
+
+## Purpose and scope
+
+This package is a commit-pinned discovery baseline and feature index for the `discovery` branch at frozen commit `d8a2059df8e53bc7275c45339849d14c8651e73c`. It inventories **79 feature records** across **16 domains** to help maintainers locate likely implementation areas and identify validation gaps. It is not an authoritative architecture reference or a runtime-certification record.
+
+The canonical inventory is [`feature-catalog.json`](feature-catalog.json); [`feature-catalog.md`](feature-catalog.md) and the files in [`domains/`](domains/) are derived reading views. See [`audit-method.md`](audit-method.md) for the status and maturity definitions.
+
+## What maintainers may use now
+
+- Use the catalog and domain views as a frozen discovery index, including their feature IDs, stated scope, likely source locations, and declared runtime prerequisites.
+- Treat a catalog status such as `verified` as meaning implementation was identified during discovery. It does **not** mean every evidence locator, line range, test claim, or runtime behaviour has passed semantic validation.
+- Use the structural checks to confirm package shape and cross-view consistency; use the evidence validator to assess whether individual evidence assertions are semantically supported.
+
+## Validation snapshot
+
+Structural validation passes: the catalog has 79 records, the 16 domain views match it, and the package structural validators pass. All 11 focused evidence-validator tests pass.
+
+The evidence validator found **170 evidence entries**: **82 valid**, **66 invalid**, **6 ambiguous**, and **16 unsupported**. Its non-zero exit is expected while those semantic evidence defects remain.
+
+Structural validity checks the documentation schema, record counts, derived-view consistency, file existence, line-range bounds, links, and public-safety rules. Semantic evidence validity additionally checks whether the cited locator exists, falls within its cited range, uses a supported parser, and actually supports the feature claim. Passing the former does not establish the latter.
+
+## E2 review decisions
+
+E2 means directly relevant automated test evidence supports the feature claim; a test file’s existence alone is insufficient. The generated evidence report was used to reassess all ten E2 records.
+
+| Feature | Decision | Reason |
+|---|---|---|
+| `CHAT-001` | Retain E2 | Two cited tests directly exercise documented streaming-related behaviour. |
+| `MODEL-006` | Retain E2 | The cited device-flow test exercises the Copilot start/poll contract and verification URI. |
+| `MODEL-007` | Retain E2 | The cited device-flow test exercises the ChatGPT subscription contract and verification URI. |
+| `RESEARCH-003` | Demote to E1 | The cited test covers ranking, not provider connectivity or dispatch; the route-to-provider implementation path was identified. |
+| `DOCUMENT-002` | Demote to E1 | The cited marker test does not support form processing or rendering; the document route does call the PDF form handlers. |
+| `EMAIL-001` | Demote to E1 | The cited health test is narrower than setup, SMTP, and polling; application setup invokes the email router and its poller. |
+| `SECURITY-002` | Demote to E1 | Relevant vault-password tests exist, but the cited test locator is fabricated; the application mounts the vault route implementation. |
+| `SECURITY-004` | Demote to E1 | Relevant injection tests exist, but the cited locator names are fabricated; callers use the documented context guard. |
+| `SECURITY-005` | Demote to E1 | Relevant guard tests exist, but the cited locator names are fabricated; route code calls the documented URL guard. |
+| `PLATFORM-009` | Demote to E0 | The manifests and diagnostic script establish discovered operational artifacts, not a traced application path or directly relevant automated test. |
+
+The current maturity distribution is **E0: 68**, **E1: 8**, **E2: 3**, **E3: 0**, **E4: 0**. Runtime validation is still pending where the catalog says it requires external services, interactive authentication, specialised hardware, or host Docker GPU support.
+
+## Known discovery caveats
+
+- `AGENT-004` includes a legacy no-op activity-log shim rather than active assistant-log behaviour.
+- `FRONTEND-007` points to a missing `static/backgrounds.html` target; its existing variant pages do not make that route functional.
+- `RESEARCH-003` retains a compatibility module that aliases the canonical search implementation, and `DOCUMENT-002` separates optional PDF viewing from form handling.
+- Secret-storage and vault-command handling are distinct implementation areas; this index does not make an end-to-end security guarantee.
+
+## Recommended next documentation work
+
+Repair semantic evidence selectively while architecture and operations documentation is written, beginning with the seven E2 demotions and maintainer-owned feature descriptions. Do not wait for the complete evidence queue before documenting the system. Record controlled runtime observations when external services, credentials, hardware, or Docker GPU access are available, and label unsupported claims explicitly.
+
+## Validation commands
+
+```bash
+PYTHONDONTWRITEBYTECODE=1 python3 docs/discovery/tools/validate_discovery_docs.py --repo-root .
+PYTHONDONTWRITEBYTECODE=1 python3 docs/discovery/tools/validate_discovery_consistency.py
+PYTHONDONTWRITEBYTECODE=1 python3 -m unittest docs/discovery/tools/test_validate_discovery_evidence.py -v
+PYTHONDONTWRITEBYTECODE=1 python3 docs/discovery/tools/validate_discovery_evidence.py --repo-root . --catalog docs/discovery/feature-catalog.json --output-dir <local-report-directory>
+```
+
+Supply a local report directory outside `docs/discovery/` for the final command so generated reports are not added to the package.

docs/discovery/README.md

@@ -0,0 +1,21 @@
+# Odysseus Discovery Package
+
+## Provisional discovery baseline
+
+This is a commit-pinned discovery baseline and feature index for the `discovery` branch at `d8a2059df8e53bc7275c45339849d14c8651e73c`. It contains 79 feature records across 16 domains. It is **not** an authoritative architecture reference, a runtime certification, or a claim that every evidence citation is semantically valid.
+
+Read [`BASELINE-STATUS.md`](BASELINE-STATUS.md) first for the publication status, evidence-validation totals, E2 decisions, known caveats, and the recommended next documentation work.
+
+## Package contents
+
+- [`feature-catalog.json`](feature-catalog.json) is the canonical machine-readable catalog.
+- [`feature-catalog.md`](feature-catalog.md) and [`domains/`](domains/) are derived reading views.
+- [`audit-method.md`](audit-method.md) defines feature status and evidence maturity.
+- [`references/source-provenance.md`](references/source-provenance.md) records the frozen snapshot.
+- [`tools/`](tools/) contains the structural, consistency, and evidence validators.
+
+A feature status such as `verified` means implementation was identified during discovery. It does not mean every evidence locator, line range, test claim, or runtime behaviour has passed semantic validation.
+
+## Validation
+
+Run the commands in [`BASELINE-STATUS.md`](BASELINE-STATUS.md#validation-commands). Structural checks and semantic evidence checks have different purposes; see that status document for the current results and interpretation.

docs/discovery/audit-method.md

@@ -0,0 +1,80 @@
+# Odysseus Functional Audit Methodology
+
+## Purpose
+
+This document specifies the methodology and evidence standards for the read-only discovery audit of **Odysseus**.
+
+## Snapshot Baseline
+
+- **Repository**: `odysseus-dev/odysseus`
+- **Audit Target Branch**: `discovery`
+- **Frozen Commit SHA**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Snapshot Date**: `2026-07-23T14:49:02Z`
+- **Audit Mode**: Read-Only inventory & documentation review
+
+## Rules of Engagement
+
+1. **No Code Mutations**: Application code and tests outside `docs/discovery/` remain untouched.
+2. **No External Operations**: No GitHub issues, PRs, comments, labels, or branch mutations.
+3. **Zero Inferred Success**: Documentation claims require empirical evidence of implementation and reachability. Filenames, README descriptions, and docstrings alone do not constitute proof.
+4. **Strict Status Categorization**: All capabilities are assigned exactly one authorized status:
+   - `verified`: Implemented, reachable, and supported by code evidence.
+   - `partial`: Partially implemented or missing full frontend/backend connection.
+   - `disabled`: Gated off by default feature flags or configuration.
+   - `experimental`: Active but requiring non-standard hardware or runtimes.
+   - `legacy`: Obsolete feature retained for backwards compatibility.
+   - `dead-code-candidate`: Code exists but is unreachable from UI or API routes.
+   - `unverified`: Implementation present but untestable without external secrets or hardware.
+
+## Evidence Maturity Scale
+
+Evidence maturity is evaluated independently from catalog feature status. A feature status such as `verified` records that implementation was identified during discovery; it is not a statement that every evidence locator, test claim, line range, or runtime behaviour has passed semantic validation.
+
+- **E0 - Discovered**: Candidate identified in documentation, route declaration, or source file.
+- **E1 - Code-path traced**: Frontend/API entry point connected through services and data handlers.
+- **E2 - Test-backed**: At least one directly relevant automated test assertion supports the feature claim. A test file's existence, an unrelated assertion, or an invalid test locator does not establish E2.
+- **E3 - Runtime-validated**: Maintainer reproduced behavior in a recorded local environment.
+- **E4 - Maintainer-accepted**: Maintainers accepted the feature description and support status.
+
+## Audit Workflow
+
+```mermaid
+flowchart TD
+    P0["Phase 0: Snapshot Isolation<br/>(Commit d8a2059)"] --> P1["Phase 1: Codebase Discovery<br/>(Routes, Services, Static JS, Specs)"]
+    P1 --> P2["Phase 2: Feature Reachability & Verification<br/>(Route matching, FE entrypoints, tests)"]
+    P2 --> P3["Phase 3: Catalog & Evidence Compilation<br/>(File paths, symbols, exact line ranges)"]
+    P3 --> P4["Phase 4: Quality & Integrity Audit<br/>(100% path existence check, schema validation)"]
+```
+
+### Phase 0: Snapshot Isolation
+The audit is pinned to git commit `d8a2059df8e53bc7275c45339849d14c8651e73c`. All file paths, symbol declarations, and line ranges map strictly to this commit.
+
+### Phase 1: Codebase Discovery
+All top-level and nested directories were traversed, including:
+- Backend Entry Points (`app.py`, `routes/`, `routes/*/*.py`, `companion/`)
+- Core Framework (`core/database.py`, `core/session_manager.py`, `core/auth.py`)
+- Business Logic Services (`src/`, `services/`, `mcp_servers/`)
+- Frontend Assets (`static/app.js`, `static/js/`, `static/index.html`)
+- Test Suites (`tests/`, `tests/cli/`, `tests/streaming/`)
+- Operations & Docker (`Dockerfile`, `docker-compose*.yml`, `scripts/`)
+
+### Phase 2: Verification Protocol
+For each feature candidate, the following table was evaluated:
+- **User Reachability**: Frontend UI element, modal, route, or CLI script.
+- **API Entrypoint**: FastAPI `@router` declaration or WebSocket/SSE handler.
+- **Backend Execution**: Concrete Python module method, service, or tool call.
+- **Data Persistence**: Disk file, SQLite table, or vector collection.
+- **Test Coverage**: Automated test file executing assertions against the component.
+
+### Phase 3: Evidence Linking Standard
+Every feature entry in `feature-catalog.json` contains a structured `evidence` list with:
+- `path`: Relative path from repository root.
+- `symbol`: Route, class, function, or element symbol name.
+- `line_range`: Inclusive line range (e.g. `L120-L250`).
+- `explanation`: Short factual statement proving reachability or implementation.
+
+### Phase 4: Quality Check & Schema Constraints
+Before finalization:
+1. Every evidence file path is validated against the checkout.
+2. Every Markdown entry matches `feature-catalog.json`.
+3. Recommendation language is separated from empirical factual observations.

docs/discovery/domains/agent.md

@@ -0,0 +1,153 @@
+# Agent
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## AGENT-001 — Autonomous Agent Loop & Tool Execution Engine
+
+- **Domain**: `agent`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Executes multi-step agent reasoning loops, tool invocation parsing, and automated response generation.
+
+### Evidence summary
+
+- `src/agent_loop.py` — `run_agent_loop` — Core loop evaluating model tool requests and executing handlers.
+- `src/tool_execution.py` — `execute_tool_call` — Dispatches tool invocation requests to underlying tool handlers.
+
+### Unknowns
+
+- Infinite tool loop if termination condition fails.
+
+## AGENT-002 — Scheduled Tasks & Event Bus Dispatcher
+
+- **Domain**: `agent`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Schedules background recurring or delayed tasks, emits event bus triggers, and executes automated flows.
+
+### Evidence summary
+
+- `routes/task_routes.py` — `@router.get('')` — Fetches active scheduled tasks.
+- `src/task_scheduler.py` — `TaskScheduler` — Async task scheduler dispatching cron and delay triggers.
+
+### Unknowns
+
+- Task execution failure handling on system restart.
+
+## AGENT-003 — Webhook Event Subscriptions & Trigger Processing
+
+- **Domain**: `agent`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Manages incoming/outgoing webhook subscriptions, endpoint authentication tokens, and event triggers.
+
+### Evidence summary
+
+- `routes/webhook_routes.py` — `@router.get('/webhooks')` — Returns list of registered webhooks.
+- `src/webhook_manager.py` — `WebhookManager` — Handles payload delivery and signature verification.
+
+### Unknowns
+
+- SSRF risks when contacting external webhook URLs if unvalidated.
+
+## AGENT-004 — Assistant Settings, Task Check-Ins & Background Job Monitor
+
+- **Domain**: `agent`
+- **Status**: `partial`
+- **Evidence Maturity**: `E1`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Manages per-user assistant sessions and scheduled check-in settings, drains background job completions, and retains a legacy no-op activity logging shim.
+
+### Evidence summary
+
+- `routes/assistant_routes.py` — `setup_assistant_routes` — Active assistant session, settings, manual check-in, run-status and timezone-list endpoints, including the owner-scoping guards.
+- `src/bg_monitor.py` — `_drain_agent` — Runs the agent loop headless against a session to produce the background-job follow-up turn.
+- `src/bg_monitor.py` — `_run_followup` — Drains completed background jobs and auto-continues the owning session, deferring while a live turn is in progress.
+- `src/assistant_log.py` — `log_to_assistant` — Legacy no-op activity logging shim retained for existing callers; documented as inactive rather than as current behaviour.
+
+### Unknowns
+
+- Route `/api/assistant/logs` cited in legacy docs is absent from assistant router.
+- Existing unit test `tests/cli/test_logs_cli_resolve_nonstring.py` tests CLI target-name resolution logic, not active assistant routes or bg_monitor execution loop.
+
+## AGENT-005 — Model Context Protocol (MCP) Server Integration
+
+- **Domain**: `agent`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Integrates external MCP servers over stdio/SSE to expand agent capabilities dynamically.
+
+### Evidence summary
+
+- `routes/mcp_routes.py` — `setup_mcp_routes` — Exposes management endpoints for external MCP servers.
+- `src/mcp_manager.py` — `McpManager` — Manages MCP server subprocess lifecycles.
+
+### Unknowns
+
+- Subprocess leaks if external MCP server process fails to terminate clean.
+
+## AGENT-006 — AI Interaction Tools & Pipeline Orchestration
+
+- **Domain**: `agent`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Provides specialized AI interaction tools for agent self-debugging, debate, and multi-model collaboration.
+
+### Evidence summary
+
+- `src/ai_interaction.py` — `init_ai_interaction_tools` — Registers specialized multi-agent interaction primitives.
+- `src/builtin_actions.py` — `execute_builtin_action` — Executes pre-built action intent sequences.
+
+### Unknowns
+
+- High API token consumption during extended agent debates.
+
+## AGENT-007 — Subprocess & Background Job Execution Tools
+
+- **Domain**: `agent`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Provides sandboxed bash/shell tool execution capabilities with output streaming and background tracking.
+
+### Evidence summary
+
+- `src/agent_tools/subprocess_tools.py` — `run_command` — Executes shell commands in background/foreground.
+- `src/bg_jobs.py` — `JobManager` — Tracks async background subprocess tasks.
+
+### Unknowns
+
+- Arbitrary shell command execution permissions if sandbox confinement fails.

docs/discovery/domains/calendar.md

@@ -0,0 +1,45 @@
+# Calendar
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## CALENDAR-001 — CalDAV Calendar Synchronization & Account Setup
+
+- **Domain**: `calendar`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: pending — Requires a controlled external CalDAV server.
+
+### Purpose
+
+Connects to remote CalDAV servers (Apple iCloud, Nextcloud, Google) to sync calendar event feeds.
+
+### Evidence summary
+
+- `routes/calendar_routes.py` — `setup_calendar_routes` — Exposes CalDAV setup and manual sync trigger routes.
+- `src/caldav_sync.py` — `CalDavSync` — Fetches and parses remote iCalendar VEVENT objects.
+
+### Unknowns
+
+- Invalid SSL certificates on self-hosted CalDAV servers.
+
+## CALENDAR-002 — Calendar Event Operations & iCalendar Parsing
+
+- **Domain**: `calendar`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Creates, updates, deletes, and displays calendar events with timezone conversion and reminder notifications.
+
+### Evidence summary
+
+- `routes/calendar_routes.py` — `@router.get('/events')` — Fetches calendar events for requested date window.
+- `src/tools/calendar.py` — `CalendarTool` — Agent tool for creating and modifying calendar entries.
+
+### Unknowns
+
+- Recurring RRULE event expansion calculation bugs across leap years.

docs/discovery/domains/chat.md

@@ -0,0 +1,197 @@
+# Chat
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## CHAT-001 — Core Chat Streaming & SSE Message Generation
+
+- **Domain**: `chat`
+- **Status**: `verified`
+- **Evidence Maturity**: `E2`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: pending — Requires access to a live LLM provider endpoint (OpenAI API key or local Ollama server).
+
+### Purpose
+
+Handles real-time Server-Sent Events (SSE) chat streaming, token rendering, and model response generation.
+
+### Evidence summary
+
+- `routes/chat_routes.py` — `chat_stream` — POST /api/chat_stream SSE endpoint; builds the shared chat context, then dispatches to the chat-mode or agent-mode streaming path.
+- `routes/chat_helpers.py` — `build_chat_context` — Shared context builder invoked by chat_stream; runs message preprocessing and assembles the memory/RAG/web context preface.
+- `src/chat_handler.py` — `ChatHandler.preprocess_message` — Message preprocessing (attachments, URLs, tool preprocessing) reached from build_chat_context via routes/chat_helpers.py:preprocess.
+- `src/chat_processor.py` — `ChatProcessor.build_context_preface` — Builds the retrieval and web-source context preface injected into the streamed request.
+- `src/llm_core.py` — `stream_llm_with_fallback` — Chat-mode streaming dispatcher called from chat_stream; wraps stream_llm with an ordered provider fallback chain.
+- `src/llm_core.py` — `stream_llm` — Per-request streaming entry wrapped by stream_llm_with_fallback; acquires the local model slot and delegates to _stream_llm_inner.
+- `src/agent_loop.py` — `stream_agent_loop` — Agent-mode streaming path called from chat_stream when the request selects agent mode.
+- `tests/test_chat_metrics.py` — `test_stream_llm_passes_through_llamacpp_timings` — Inspected unit test asserting stream_llm forwards backend generation timings into the emitted metrics chunk.
+- `tests/test_resend_message_nondestructive.py` — `test_resend_message_does_not_truncate_by_default` — Inspected unit test asserting the frontend resend path does not truncate prior conversation turns.
+
+### Unknowns
+
+- Stream interruption on connection drops requires retry logic.
+
+## CHAT-002 — Session Management & Conversation State
+
+- **Domain**: `chat`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Manages session creation, listing, switching, renaming, and persistence of conversation metadata.
+
+### Evidence summary
+
+- `routes/session_routes.py` — `@router.get('/api/sessions')` — Lists active sessions filtered by user owner scope.
+- `core/session_manager.py` — `SessionManager` — Provides thread-safe session storage operations.
+
+### Unknowns
+
+- Concurrent file writes to sessions.json under high load.
+
+## CHAT-003 — Chat History & Message Editing/Truncation
+
+- **Domain**: `chat`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Provides history retrieval, message content updating, message deletion, and history branch truncation.
+
+### Evidence summary
+
+- `routes/history/history_routes.py` — `@router.get('/api/history/{session_id}')` — Fetches message history timeline for a session.
+- `routes/history_routes.py` — `_sys.modules[__name__] = _canonical` — Backward-compatibility shim module.
+
+### Unknowns
+
+- Truncating messages re-indexes context window and clears cached tool calls.
+
+## CHAT-004 — File & Multimodal Attachment Handling
+
+- **Domain**: `chat`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Handles uploading, mime validation, image preview, vision encoding, and file attachments in chat messages.
+
+### Evidence summary
+
+- `routes/upload_routes.py` — `@router.post('')` — Accepts multi-part file uploads and generates vision metadata.
+- `src/upload_handler.py` — `UploadHandler.save_file` — Validates upload size and atomicity on disk.
+
+### Unknowns
+
+- Large file uploads may consume server disk space if cleanup task fails.
+
+## CHAT-005 — Chat Message Search
+
+- **Domain**: `chat`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Enables full-text keyword search across stored chat messages and sessions.
+
+### Evidence summary
+
+- `routes/search_routes.py` — `setup_search_routes` — Registers chat message search endpoint.
+- `src/session_search.py` — `search_sessions` — Executes query matching against session transcripts.
+
+### Unknowns
+
+- Full table scans on un-indexed text columns for very large databases.
+
+## CHAT-006 — System Prompts & Preset Management
+
+- **Domain**: `chat`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Provides creation, selection, and customization of system prompt presets for chat sessions.
+
+### Evidence summary
+
+- `routes/preset_routes.py` — `setup_preset_routes` — API routes for listing and modifying system prompt presets.
+- `src/preset_manager.py` — `PresetManager` — Disk-backed manager for prompt presets.
+
+### Unknowns
+
+- Invalid JSON syntax in user presets file can corrupt preset loading.
+
+## CHAT-007 — Emoji Rendering & Twemoji SVG Proxy
+
+- **Domain**: `chat`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Proxies Twemoji SVG icons locally to render flat SVG emojis in message text without external CDN dependencies.
+
+### Evidence summary
+
+- `routes/emoji_routes.py` — `setup_emoji_routes` — Serves locally cached Twemoji SVGs.
+
+### Unknowns
+
+- First request fetches SVG from remote CDN before caching locally.
+
+## CHAT-008 — Input History Recall (Arrow Up)
+
+- **Domain**: `chat`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Allows users to cycle through previously sent prompt messages in the chat composer input using Arrow-Up/Down keys.
+
+### Evidence summary
+
+- `static/js/composerArrowUpRecall.js` — `initComposerRecall` — Listens for ArrowUp keypress on composer textarea.
+
+### Unknowns
+
+- Client-side browser storage limits.
+
+## CHAT-009 — Context Window Compaction & Truncation
+
+- **Domain**: `chat`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Compacts session transcript history when prompt size exceeds context limits using summarization.
+
+### Evidence summary
+
+- `routes/history/history_routes.py` — `@router.post('/api/session/{session_id}/compact')` — Triggers context summarization and compaction.
+- `src/context_compactor.py` — `compact_context` — Executes context token pruning and summary generation.
+
+### Unknowns
+
+- Aggressive compaction may discard subtle user instructions.

docs/discovery/domains/contact.md

@@ -0,0 +1,24 @@
+# Contact
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## CONTACT-001 — CardDAV Contact Management & Address Book Integration
+
+- **Domain**: `contact`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Connects to CardDAV servers, imports VCard contacts, and provides contact lookup for email/calendar autocomplete.
+
+### Evidence summary
+
+- `routes/contacts/contacts_routes.py` — `@router.get('/list')` — Returns contact list filtered by search query.
+- `src/tools/contacts.py` — `ContactsTool` — Agent tool for querying user address book contacts.
+
+### Unknowns
+
+- VCard 3.0 vs 4.0 property parsing mismatches.

docs/discovery/domains/cookbook.md

@@ -0,0 +1,85 @@
+# Cookbook
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## COOKBOOK-001 — Local Model Download & Recipe Lifecycle Management
+
+- **Domain**: `cookbook`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Downloads HuggingFace models, configures execution parameters, and manages local GGUF/MLX model servers.
+
+### Evidence summary
+
+- `routes/cookbook_routes.py` — `setup_cookbook_routes` — Exposes model downloading and process serving endpoints.
+- `static/js/cookbook.js` — `initCookbook` — UI manager for local model library.
+
+### Unknowns
+
+- Disk space exhaustion during multi-gigabyte GGUF weights downloads.
+
+## COOKBOOK-002 — Hardware Model Fitting ('What Fits?') Analysis Engine
+
+- **Domain**: `cookbook`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Calculates RAM/VRAM requirements, quantized size, and context overhead to determine model compatibility.
+
+### Evidence summary
+
+- `routes/hwfit_routes.py` — `setup_hwfit_routes` — Calculates hardware model compatibility.
+- `services/hwfit/fit.py` — `calculate_fit` — Performs parameter and memory fit calculations.
+
+### Unknowns
+
+- Inaccurate VRAM estimation for non-standard KV-cache quantization.
+
+## COOKBOOK-003 — HuggingFace & MLX Model Discovery Services
+
+- **Domain**: `cookbook`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Searches HuggingFace Hub and local MLX model repositories for compatible GGUF and MLX weights.
+
+### Evidence summary
+
+- `services/hwfit/hf_discovery.py` — `search_hf_models` — Queries HuggingFace API for model tags and files.
+
+### Unknowns
+
+- HuggingFace API rate limits when searching without an API token.
+
+## COOKBOOK-004 — Host Docker Access for Inference Container Runtimes
+
+- **Domain**: `cookbook`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: pending — Requires Docker access and supported physical GPU hardware.
+
+### Purpose
+
+Detects and connects to host Docker engine to launch containerized Ollama, vLLM, or SGLang runtimes.
+
+### Evidence summary
+
+- `src/host_docker_access.py` — `HostDockerAccess` — Interacts with host docker daemon.
+
+### Unknowns
+
+- Permission denied accessing docker socket on non-root setups.

docs/discovery/domains/document.md

@@ -0,0 +1,108 @@
+# Document
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## DOCUMENT-001 — Document & Canvas Artifact Management
+
+- **Domain**: `document`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Renders dynamic canvas documents, handles live editing, markdown preview, and side-by-side artifact display.
+
+### Evidence summary
+
+- `routes/document_routes.py` — `setup_document_routes` — Registers document artifact CRUD routes.
+- `static/js/document.js` — `initDocumentView` — Renders interactive canvas document panel.
+
+### Unknowns
+
+- Concurrent edits on the same document artifact.
+
+## DOCUMENT-002 — PDF Form Processing & High-Fidelity Rendering
+
+- **Domain**: `document`
+- **Status**: `verified`
+- **Evidence Maturity**: `E1`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: pending — Requires optional PyMuPDF (`fitz`) or pypdf runtime dependency.
+
+### Purpose
+
+Extracts form fields from PDF files, fills dynamic values, and generates PDF previews.
+
+### Evidence summary
+
+- `src/pdf_runtime.py` — `load_pymupdf_for_pdf_viewer` — Loads optional PyMuPDF runtime for PDF viewing.
+- `src/pdf_forms.py` — `extract_form_fields` — Handles PDF form field extraction and filling.
+- `tests/test_document_pdf_marker.py` — `test_marker_removed_without_eating_following_text` — Tests PDF text extraction wrapper stripping without content corruption.
+
+### Unknowns
+
+- Complex XFA PDF forms may not extract cleanly with standard pdf parsers.
+
+## DOCUMENT-003 — Personal Document Indexing & RAG Retrieval
+
+- **Domain**: `document`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Indexes local user documents (PDF, DOCX, TXT) into ChromaDB for semantic vector retrieval.
+
+### Evidence summary
+
+- `routes/personal_routes.py` — `setup_personal_routes` — Personal document RAG indexing and search API endpoints.
+- `src/personal_docs.py` — `PersonalDocsManager` — Handles file text chunking and vector storage.
+
+### Unknowns
+
+- Slow vector embedding indexing step for massive multi-thousand page documents.
+
+## DOCUMENT-004 — Document Conversion & Text Extraction Engine
+
+- **Domain**: `document`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Converts office formats (.docx, .xlsx, .pptx) and HTML into clean Markdown text representations.
+
+### Evidence summary
+
+- `src/markitdown_runtime.py` — `convert_to_markdown` — Converts binary office documents into structured Markdown text.
+
+### Unknowns
+
+- Formatting loss when parsing legacy binary doc/xls files.
+
+## DOCUMENT-005 — Document Library UI Navigation
+
+- **Domain**: `document`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Provides dedicated UI view for browsing, filtering, and organizing saved user documents.
+
+### Evidence summary
+
+- `static/js/documentLibrary.js` — `initDocumentLibrary` — Renders document library navigation grid.
+- `app.py` — `serve_library` — Serves SPA shell for /library route.
+
+### Unknowns
+
+- Large folder trees may cause initial DOM render slowdown.

docs/discovery/domains/email.md

@@ -0,0 +1,87 @@
+# Email
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## EMAIL-001 — Email Account Setup, IMAP/SMTP Connection & Polling
+
+- **Domain**: `email`
+- **Status**: `verified`
+- **Evidence Maturity**: `E1`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: pending — Requires a controlled live IMAP account and network access.
+
+### Purpose
+
+Configures IMAP/SMTP email accounts, validates TLS certificates, and polls background inbox updates.
+
+### Evidence summary
+
+- `routes/email_routes.py` — `setup_email_routes` — Sets up email account management and synchronization routes.
+- `routes/email_pollers.py` — `_start_poller` — Background poller for email inbox synchronization.
+- `tests/test_service_health_email.py` — `test_email_ok_all_connect` — Tests IMAP connection health probing and status reporting.
+
+### Unknowns
+
+- Account lockouts if bad credentials are repeatedly polled.
+
+## EMAIL-002 — Email Searching, Threading & Message Operations
+
+- **Domain**: `email`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Parses email headers, folds signatures, groups messages into threads, and executes full-text email search.
+
+### Evidence summary
+
+- `routes/email_routes.py` — `@router.get('/search')` — Executes search across cached email headers and text.
+- `src/email_thread_parser.py` — `parse_email_thread` — Builds conversation tree from Message-ID and In-Reply-To headers.
+
+### Unknowns
+
+- Malformed MIME email structures failing HTML sanitization.
+
+## EMAIL-003 — Email Composition, Draft Management & Sending
+
+- **Domain**: `email`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: pending — Requires a controlled live SMTP account and network access.
+
+### Purpose
+
+Creates, saves, and dispatches HTML/plaintext email messages via SMTP.
+
+### Evidence summary
+
+- `routes/email_routes.py` — `@router.post('/send')` — Sends email message via user SMTP credentials.
+
+### Unknowns
+
+- SMTP connection drop mid-send causing unsent mail state.
+
+## EMAIL-004 — Email MCP Server & Codex Integration Bridge
+
+- **Domain**: `email`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Exposes constrained email reading and draft capabilities to external Codex / MCP agents with scope checks.
+
+### Evidence summary
+
+- `mcp_servers/email_server.py` — `EmailMcpServer` — MCP server exposing email tools over stdio/SSE.
+- `routes/codex_routes.py` — `setup_codex_routes` — Bridge endpoints for external Codex plugin integration.
+
+### Unknowns
+
+- Unauthorized mail sending if token scopes are improperly scoped.

docs/discovery/domains/frontend.md

@@ -0,0 +1,149 @@
+# Frontend
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## FRONTEND-001 — Single Page Application Shell & Client Router
+
+- **Domain**: `frontend`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Main HTML5 SPA shell, DOM lifecycle initializers, tab navigation, and deep-link route handlers.
+
+### Evidence summary
+
+- `static/index.html` — `index.html` — Main SPA entry point containing modal roots and CSS bundles.
+- `app.py` — `serve_index` — Serves index.html with dynamically generated CSP nonces.
+
+### Unknowns
+
+- Stale browser static cache if asset hashing is omitted during deployment.
+
+## FRONTEND-002 — Dynamic Theme, Color System & Custom Fonts
+
+- **Domain**: `frontend`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Supports dark/light themes, custom CSS variables, color picker controls, and user font uploads.
+
+### Evidence summary
+
+- `static/js/theme.js` — `applyTheme` — Applies custom HSL theme variables to DOM document root.
+- `routes/font_routes.py` — `setup_font_routes` — Allows uploading and serving custom WOFF2 font files.
+
+### Unknowns
+
+- Flash of unstyled content (FOUC) on slow connections.
+
+## FRONTEND-003 — Window Manager, Tile Layout & Modal Control System
+
+- **Domain**: `frontend`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Manages draggable tool windows, snapped multi-tile viewports, modal dialog Z-ordering, and ESC key stacks.
+
+### Evidence summary
+
+- `static/js/modalManager.js` — `ModalManager` — Controls modal open/close transitions and focus trapping.
+- `static/js/tileManager.js` — `TileManager` — Handles viewport split-pane grid arrangements.
+
+### Unknowns
+
+- Overlap artifacts when opening many simultaneous tool floating windows.
+
+## FRONTEND-004 — Global Keyboard Shortcuts & Accessibility Controls
+
+- **Domain**: `frontend`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Provides configurable hotkeys (Ctrl+K search, Esc close, Alt+1-9 tabs) and high-contrast accessibility options.
+
+### Evidence summary
+
+- `static/js/keyboard-shortcuts.js` — `initShortcuts` — Binds global keydown handlers for system shortcuts.
+- `static/js/a11y.js` — `initA11y` — Applies ARIA roles and dyslexic font toggles.
+
+### Unknowns
+
+- Browser keybinding collisions with browser default hotkeys.
+
+## FRONTEND-005 — Markdown, LaTeX & Code Block Streaming Renderer
+
+- **Domain**: `frontend`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Parses incoming SSE markdown streams, renders KaTeX math formulas, syntax-highlighted code, and interactive runners.
+
+### Evidence summary
+
+- `static/js/markdown.js` — `renderMarkdown` — Converts markdown prose to HTML nodes with syntax highlighting.
+- `static/js/streamingSegmenter.js` — `Segmenter` — Parses un-closed markdown fences during live stream.
+
+### Unknowns
+
+- DOM thrashing if streaming segmenter updates UI too frequently.
+
+## FRONTEND-006 — Interactive Tour & Guided Onboarding System
+
+- **Domain**: `frontend`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Presents interactive step-by-step feature tours and UI tooltip hints for new users.
+
+### Evidence summary
+
+- `static/js/tourHints.js` — `startTour` — Renders guided feature tour overlays over target UI elements.
+
+### Unknowns
+
+- Tour step misalignment if window is resized mid-tour.
+
+## FRONTEND-007 — Background Effects Prototyping Sandbox
+
+- **Domain**: `frontend`
+- **Status**: `dead-code-candidate`
+- **Evidence Maturity**: `E1`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Standalone sandbox page for prototyping visual background animations, waves, and whirlpool effects.
+
+### Evidence summary
+
+- `app.py` — `serve_backgrounds` — Serves visual background sandbox HTML page route.
+- `static/wave-variants.html` — `wave-variants.html` — Interactive background effect prototyping sandbox variant.
+
+### Unknowns
+
+- Route `/backgrounds` in app.py L918 attempts to serve `static/backgrounds.html` which is missing from disk; variant templates `wave-variants.html` and `whirlpool-variants.html` exist.

docs/discovery/domains/media.md

@@ -0,0 +1,170 @@
+# Media
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## MEDIA-001 — Gallery Image Library & Album Operations
+
+- **Domain**: `media`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Organizes images into custom albums, provides grid browsing, tagging, and album metadata management.
+
+### Evidence summary
+
+- `routes/gallery/gallery_routes.py` — `@router.get('/api/gallery/library')` — Fetches image library list with tag filters.
+- `static/js/gallery.js` — `initGallery` — Main gallery grid renderer and uploader.
+
+### Unknowns
+
+- Thumbnail generation overhead for high-resolution RAW camera images.
+
+## MEDIA-002 — Image Processing, AI Upscaling & Style Transfer
+
+- **Domain**: `media`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Executes local image enhancement, background removal, face sharpening, and AI upscaling operations.
+
+### Evidence summary
+
+- `routes/gallery/gallery_routes.py` — `@router.post('/api/gallery/ai-upscale')` — Runs RealESRGAN image upscaling.
+- `routes/gallery/gallery_routes.py` — `@router.post('/api/image/remove-bg')` — Executes background removal pass.
+
+### Unknowns
+
+- High GPU memory allocation when upscaling 4K images.
+
+## MEDIA-003 — Interactive Image Canvas Editor & Persisted Drafts
+
+- **Domain**: `media`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Provides full multi-layer raster canvas editor, brush tools, transforms, masks, and draft project persistence.
+
+### Evidence summary
+
+- `routes/editor_draft_routes.py` — `setup_editor_draft_routes` — API routes for saving and loading canvas project drafts.
+- `static/js/editor/history-panel.js` — `HistoryManager` — Canvas undo/redo stack manager.
+
+### Unknowns
+
+- Browser memory leak if multi-gigabyte layer undo buffers are kept indefinitely.
+
+## MEDIA-004 — Text-to-Speech (TTS) Synthesis Service
+
+- **Domain**: `media`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Synthesizes spoken audio from text using local Kokoro, EdgeTTS, or OpenAI TTS engines.
+
+### Evidence summary
+
+- `routes/tts_routes.py` — `@router.post('/synthesize')` — Synthesizes TTS audio clip.
+- `services/tts/tts_service.py` — `TTSService` — Provider abstraction layer for audio speech generation.
+
+### Unknowns
+
+- Audio synthesis latency on CPU-only hardware setups.
+
+## MEDIA-005 — Speech-to-Text (STT) Audio Transcription Service
+
+- **Domain**: `media`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Transcribes user audio recordings into text using faster-whisper or local speech models.
+
+### Evidence summary
+
+- `routes/stt_routes.py` — `@router.post('/transcribe')` — Accepts multipart audio file and returns transcription text.
+- `services/stt/stt_service.py` — `STTService` — Whisper audio transcription engine wrapper.
+
+### Unknowns
+
+- Missing ffmpeg system dependency prevents audio format decoding.
+
+## MEDIA-006 — Digital Signature Stamp Storage & Placement
+
+- **Domain**: `media`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Stores transparent PNG user signatures and stamps for placement onto PDF forms and documents.
+
+### Evidence summary
+
+- `routes/signature_routes.py` — `setup_signature_routes` — CRUD endpoints for managing user signature PNG stamps.
+
+### Unknowns
+
+- Cross-site scripting if signature image titles contain unescaped user input.
+
+## MEDIA-007 — Generated Image Artifact Route & MCP Integration
+
+- **Domain**: `media`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Serves generated AI artwork artifacts and integrates with image generation MCP server.
+
+### Evidence summary
+
+- `app.py` — `serve_generated_image` — Serves generated image artifacts with cache headers.
+- `src/generated_images.py` — `resolve_generated_image_path` — Confines requested image path within artifacts directory.
+
+### Unknowns
+
+- Path traversal vulnerability if filename parameter is un-sanitized.
+
+## MEDIA-008 — Native MLX Image Bridge (macOS Apple Silicon)
+
+- **Domain**: `media`
+- **Status**: `experimental`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: pending — Requires Apple Silicon, macOS tooling, and the compiled MLX bridge.
+
+### Purpose
+
+Native Apple Swift bridge for hardware-accelerated diffusion and MLX image colorization on macOS.
+
+### Evidence summary
+
+- `swift/odysseus-mlx-image-bridge/Package.swift` — `Package` — Swift package manifest for native MLX image bridge.
+- `scripts/mlx_image_server.py` — `main` — Python daemon wrapping native Swift MLX binary.
+
+### Unknowns
+
+- Binary build requires Xcode command line tools build step (`build-macos-app.sh`).

docs/discovery/domains/memory.md

@@ -0,0 +1,24 @@
+# Memory
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## MEMORY-001 — Persistent Long-Term Memory & Vector Indexing
+
+- **Domain**: `memory`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Extracts facts, user preferences, and temporal memories from chat sessions into vector/relational storage.
+
+### Evidence summary
+
+- `routes/memory/memory_routes.py` — `@router.get('')` — Fetches long-term user memory timeline.
+- `services/memory/memory_extractor.py` — `MemoryExtractor` — LLM-driven fact extraction from conversation transcripts.
+
+### Unknowns
+
+- Conflicting memory facts extracted from contradictory user prompts.

docs/discovery/domains/model.md

@@ -0,0 +1,177 @@
+# Model
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## MODEL-001 — Multi-Provider LLM Model Discovery & Metadata Management
+
+- **Domain**: `model`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Discovers models from OpenAI, Anthropic, Ollama, vLLM, LMStudio, OpenRouter, and Google AI Studio endpoints.
+
+### Evidence summary
+
+- `routes/model_routes.py` — `@router.get('/api/models')` — Returns unified list of available models across providers.
+- `src/model_discovery.py` — `ModelDiscovery.discover_all` — Queries connected provider endpoints for available model IDs.
+
+### Unknowns
+
+- Remote endpoint timeouts may slow down full discovery refresh.
+
+## MODEL-002 — Model Capability & Context Limits Detection
+
+- **Domain**: `model`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Detects vision, tool calling, reasoning, and context window limits for connected model endpoints.
+
+### Evidence summary
+
+- `src/model_capabilities.py` — `get_model_capabilities` — Maps model names to vision and tool support flags.
+- `src/endpoint_resolver.py` — `resolve_endpoint_headers` — Resolves auth headers and target URLs for model endpoints.
+
+### Unknowns
+
+- Incorrect context limit metadata for unlisted custom fine-tunes.
+
+## MODEL-003 — LLM Core Provider Communication & Fallback Routing
+
+- **Domain**: `model`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: pending — Provider dispatch, header injection and fallback advancement are only observable against a reachable LLM provider endpoint; not exercised in this documentation pass.
+
+### Purpose
+
+Manages HTTP request dispatching, authorization header injection, and fallback provider routing for LLM calls.
+
+### Evidence summary
+
+- `src/llm_core.py` — `llm_call_async` — Non-streaming provider request dispatcher: resolves the endpoint, injects authorization headers and executes the HTTP call.
+- `src/llm_core.py` — `llm_call_async_with_fallback` — Ordered fallback wrapper that retries llm_call_async across the configured candidate endpoints.
+- `src/llm_core.py` — `stream_llm_with_fallback` — Ordered fallback wrapper for the streaming path; advances to the next candidate when a provider yields an empty completion.
+
+### Unknowns
+
+- Unexpected API changes in upstream third-party model providers.
+
+## MODEL-004 — Model Selection & Display Ordering Preferences
+
+- **Domain**: `model`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Allows pinning, sorting, and hiding specific models in the UI selection dropdown.
+
+### Evidence summary
+
+- `routes/model_routes.py` — `@router.post('/order')` — Saves custom model display order preference.
+
+### Unknowns
+
+- Stale model IDs in custom order lists after model endpoints are removed.
+
+## MODEL-005 — Side-by-Side Model Comparison (A/B Testing)
+
+- **Domain**: `model`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Enables dual-model side-by-side response evaluation, arena scoring, and latency comparison.
+
+### Evidence summary
+
+- `routes/compare/compare_routes.py` — `@router.post('/start')` — Starts a parallel dual-model comparison stream.
+- `static/js/compare/index.js` — `initCompareView` — Renders side-by-side model chat panes.
+
+### Unknowns
+
+- High memory and network usage when streaming two model responses simultaneously.
+
+## MODEL-006 — GitHub Copilot Device Flow Authentication
+
+- **Domain**: `model`
+- **Status**: `verified`
+- **Evidence Maturity**: `E2`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: pending — Requires an interactive GitHub Copilot OAuth device-flow account.
+
+### Purpose
+
+Authenticates with GitHub Copilot via OAuth device flow to use Copilot models directly.
+
+### Evidence summary
+
+- `routes/copilot_routes.py` — `setup_copilot_routes` — Builds the Copilot device-flow router at prefix /api/copilot, wiring _start_device_flow and _poll_device_flow.
+- `routes/device_flow.py` — `create_device_flow_router` — Shared factory registering POST /device/start and POST /device/poll under the caller-supplied prefix.
+- `src/copilot.py` — `request_device_code` — Issues the GitHub device-code request that begins the Copilot OAuth device flow.
+- `src/copilot.py` — `poll_access_token` — Polls GitHub for the access token once the user has authorized the device code.
+- `tests/test_provider_device_flow_js.py` — `test_copilot_success_uses_complete_verification_uri` — Inspected unit test asserting the Copilot device-flow runner surfaces the complete verification URI returned by the backend.
+
+### Unknowns
+
+- Token expiration requires manual device re-authentication.
+
+## MODEL-007 — ChatGPT Subscription Device Flow Authentication
+
+- **Domain**: `model`
+- **Status**: `verified`
+- **Evidence Maturity**: `E2`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: pending — Requires an interactive ChatGPT subscription OAuth flow.
+
+### Purpose
+
+Authenticates with ChatGPT Pro/Plus subscription tokens via device login flow.
+
+### Evidence summary
+
+- `routes/chatgpt_subscription_routes.py` — `setup_chatgpt_subscription_routes` — Builds the ChatGPT subscription device-flow router at prefix /api/chatgpt-subscription.
+- `routes/device_flow.py` — `create_device_flow_router` — Shared factory registering POST /device/start and POST /device/poll under the caller-supplied prefix.
+- `src/chatgpt_subscription.py` — `request_device_code` — Issues the ChatGPT device-authorization request that begins the subscription OAuth device flow.
+- `src/chatgpt_subscription.py` — `poll_device_auth` — Polls the ChatGPT device-authorization endpoint for completion using the stored device_auth_id and user_code.
+- `tests/test_provider_device_flow_js.py` — `test_chatgpt_success_uses_plain_verification_uri` — Inspected unit test asserting the ChatGPT device-flow runner uses the plain verification URI rather than the Copilot complete-URI form.
+
+### Unknowns
+
+- Changes in OpenAI auth endpoint security challenges.
+
+## MODEL-008 — Embedding Model Lane & Vector Provider Setup
+
+- **Domain**: `model`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Configures local sentence-transformers, FastEmbed, or remote OpenAI embedding model lanes.
+
+### Evidence summary
+
+- `routes/embedding_routes.py` — `setup_embedding_routes` — Provides embedding provider configuration endpoints.
+- `src/embeddings.py` — `EmbeddingManager` — Generates dense vector embeddings for RAG and memory.
+
+### Unknowns
+
+- First-time download of heavy PyTorch model weights on CPU-only machines.

docs/discovery/domains/note.md

@@ -0,0 +1,24 @@
+# Note
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## NOTE-001 — Interactive Notes & Checklist Management
+
+- **Domain**: `note`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Provides Google Keep-style notes, rich markdown text, checklist items, pinning, color tags, and reminders.
+
+### Evidence summary
+
+- `routes/note/note_routes.py` — `@router.get('')` — Lists all user notes with pin and archive states.
+- `static/js/notes.js` — `initNotesView` — Main interactive notes grid and modal manager.
+
+### Unknowns
+
+- Concurrent edits on note item checkboxes.

docs/discovery/domains/platform.md

@@ -0,0 +1,213 @@
+# Platform
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## PLATFORM-001 — Application Initialization & Lifespan Management
+
+- **Domain**: `platform`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Orchestrates server startup, database table migration, background daemon initialization, and clean shutdown.
+
+### Evidence summary
+
+- `app.py` — `_lifespan` — FastAPI lifespan context manager executing startup tasks.
+- `src/app_initializer.py` — `initialize_app` — Initializes app directories, DB schemas, and logging.
+
+### Unknowns
+
+- Un-handled exceptions during startup halt application launch.
+
+## PLATFORM-002 — System Health, Readiness & Version Monitoring APIs
+
+- **Domain**: `platform`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Exposes Liveness (/api/health), Readiness (/api/ready), App Version (/api/version), and Client Perf APIs.
+
+### Evidence summary
+
+- `app.py` — `readiness_check` — Performs system component integrity check.
+- `src/readiness.py` — `check_readiness` — Checks database, storage, and key paths for read/write access.
+
+### Unknowns
+
+- Readiness check delays if verifying connectivity to offline remote endpoints.
+
+## PLATFORM-003 — Database Schema, Migrations & SQLite Persistence
+
+- **Domain**: `platform`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Defines core relational tables (users, tokens, tasks, sessions) and executes automated SQLite schema upgrades.
+
+### Evidence summary
+
+- `core/database.py` — `init_db` — Creates ORM tables and establishes connection pool.
+- `scripts/update_database.py` — `run_migrations` — Applies missing schema columns and indices.
+
+### Unknowns
+
+- SQLite file lock contention under high concurrent write loads.
+
+## PLATFORM-004 — User Data Export & Import Backup Infrastructure
+
+- **Domain**: `platform`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Exports complete user workspace state (sessions, memory, skills, notes, presets) into a zip archive.
+
+### Evidence summary
+
+- `routes/backup_routes.py` — `setup_backup_routes` — Handles workspace data export and import upload unpack.
+- `docs/backup-restore.md` — `Documentation` — Backup and restore operational documentation.
+
+### Unknowns
+
+- Corrupt archive files causing partial data restore.
+
+## PLATFORM-005 — File Cleanup & Storage Maintenance Engine
+
+- **Domain**: `platform`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Scans data directories for orphaned files, old uploads, temporary vision images, and frees disk space.
+
+### Evidence summary
+
+- `routes/cleanup/cleanup_routes.py` — `@router.get('/preview')` — Previews reclaimable disk space across storage directories.
+- `src/cleanup_service.py` — `CleanupService` — Executes filesystem purge of orphaned asset files.
+
+### Unknowns
+
+- Deletes files uploaded in active sessions if retention window is set too short.
+
+## PLATFORM-006 — System Health & RAG Diagnostic Suite
+
+- **Domain**: `platform`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Executes real-time integrity diagnostics across ChromaDB, SearXNG, local models, and network interfaces.
+
+### Evidence summary
+
+- `routes/diagnostics_routes.py` — `setup_diagnostics_routes` — Runs subsystem health check suite.
+- `src/service_health.py` — `collect_health_status` — Inspects vector database, email, search, and local provider status.
+
+### Unknowns
+
+- Diagnostic timeout if external search provider is unreachable.
+
+## PLATFORM-007 — Desktop CLI Utilities & Shell Integration Tools
+
+- **Domain**: `platform`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Provides command-line interface tools (`odysseus`, `odysseus-mcp`, `odysseus-mail`) for terminal usage.
+
+### Evidence summary
+
+- `scripts/_lib/cli.py` — `main` — Shared CLI framework for terminal helper commands.
+- `scripts/odysseus` — `odysseus` — Main terminal launcher script.
+
+### Unknowns
+
+- Outdated CLI scripts if backend API schemas change.
+
+## PLATFORM-008 — Desktop Companion App Integration
+
+- **Domain**: `platform`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Provides API routes and pairing mechanisms for the native macOS/desktop menu bar companion app.
+
+### Evidence summary
+
+- `companion/routes.py` — `setup_companion_routes` — Endpoints for pairing and status sync with desktop companion.
+- `companion/pairing.py` — `PairingManager` — Generates and validates companion pairing codes.
+
+### Unknowns
+
+- Pairing code expiration timing window.
+
+## PLATFORM-009 — Docker Containerization & GPU Hardware Manifests
+
+- **Domain**: `platform`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: pending — Requires Docker GPU pass-through and compatible host drivers.
+
+### Purpose
+
+Provides multi-stage Dockerfile and Docker Compose manifests for CPU, NVIDIA CUDA, and AMD ROCm GPUs.
+
+### Evidence summary
+
+- `Dockerfile` — `multi-stage-build` — Multi-stage container build environment.
+- `docker-compose.gpu-nvidia.yml` — `nvidia-gpu-manifest` — NVIDIA GPU pass-through container specification.
+- `scripts/check-docker-gpu.sh` — `check-docker-gpu` — Automated diagnostic test script for host NVIDIA GPU passthrough.
+
+### Unknowns
+
+- Driver version incompatibility with host NVIDIA/AMD kernel drivers.
+
+## PLATFORM-010 — Legacy FAISS Vector Index Migration Script
+
+- **Domain**: `platform`
+- **Status**: `legacy`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Legacy utility script to migrate older FAISS vector indices into ChromaDB.
+
+### Evidence summary
+
+- `scripts/migrate_faiss_to_chroma.py` — `migrate_faiss` — Reads FAISS vector index files and writes to ChromaDB collection.
+
+### Unknowns
+
+- Superseded by native ChromaDB vector index pipeline.

docs/discovery/domains/research.md

@@ -0,0 +1,86 @@
+# Research
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## RESEARCH-001 — Deep Research Execution Engine & SSE Progress Streaming
+
+- **Domain**: `research`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Executes multi-step recursive deep research tasks, web page scraping, synthesis, and streams live progress.
+
+### Evidence summary
+
+- `routes/research/research_routes.py` — `@router.post('/api/research/start')` — Initiates deep research job.
+- `src/deep_research.py` — `DeepResearchEngine` — Recursive search and summary crawler.
+
+### Unknowns
+
+- High memory consumption when parsing multi-megabyte HTML target pages.
+
+## RESEARCH-002 — Research Library, Detail View & Image Controls
+
+- **Domain**: `research`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Stores completed research reports, generated diagrams, reference links, and manages image visibility.
+
+### Evidence summary
+
+- `routes/research/research_routes.py` — `@router.get('/api/research/library')` — Returns all saved research reports.
+
+### Unknowns
+
+- Orphaned report files if storage directory is modified out-of-band.
+
+## RESEARCH-003 — Web Search Engine Integration (SearXNG & Multi-Provider)
+
+- **Domain**: `research`
+- **Status**: `verified`
+- **Evidence Maturity**: `E1`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: pending — Requires an active SearXNG instance or external search API provider.
+
+### Purpose
+
+Queries SearXNG, DuckDuckGo, or Google Search instances to retrieve web search snippets.
+
+### Evidence summary
+
+- `routes/search_routes.py` — `setup_search_routes` — Defines /api/search, /api/search/config, and /api/search/query endpoints.
+- `src/search/core.py` — `SearchEngine` — Compatibility module aliasing services.search.core.
+- `tests/test_search_ranking.py` — `test_news_queries_prefer_news_sources_over_sports_and_social_results` — Tests search result domain ranking and scoring.
+
+### Unknowns
+
+- Search provider IP throttling or rate-limiting.
+
+## RESEARCH-004 — Research Result Peeking & Topic Spinoff Generation
+
+- **Domain**: `research`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Extracts preliminary research snippets and spawns child research sessions focused on specific sub-topics.
+
+### Evidence summary
+
+- `routes/research/research_routes.py` — `@router.post('/api/research/spinoff/{session_id}')` — Spawns child research session for specific query.
+
+### Unknowns
+
+- Deep recursion tree depth when spawning multiple nested spinoffs.

docs/discovery/domains/security.md

@@ -0,0 +1,154 @@
+# Security
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## SECURITY-001 — Authentication, Session Cookies & User Management
+
+- **Domain**: `security`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Handles bcrypt password hashing, session cookie issuance, authentication enforcement, and user administration.
+
+### Evidence summary
+
+- `routes/auth_routes.py` — `@router.post('/login')` — Authenticates credentials and sets session cookie.
+- `core/auth.py` — `AuthManager` — Handles user creation, password verification, and session tokens.
+
+### Unknowns
+
+- Cookie session hijack if deployed over unencrypted HTTP without HTTPS cookie flags.
+
+## SECURITY-002 — System Vault Encrypted Secret Storage
+
+- **Domain**: `security`
+- **Status**: `verified`
+- **Evidence Maturity**: `E1`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: pending — Requires installed Bitwarden CLI (`bw`) executable.
+
+### Purpose
+
+Encrypts API keys, passphrases, and third-party secrets on disk using AES-GCM / PBKDF2 key derivation.
+
+### Evidence summary
+
+- `routes/vault_routes.py` — `setup_vault_routes` — Admin routes for vault configuration, login, unlock, lock, and logout.
+- `src/secret_storage.py` — `SecretStorage` — Fernet symmetric key DB secret encryption.
+- `tests/test_vault_password_not_in_argv.py` — `test_bw_password_not_in_argv` — Verifies master password is fed via stdin and never appears in process argv.
+
+### Unknowns
+
+- Loss of vault master passphrase renders all encrypted secrets permanently unrecoverable.
+
+## SECURITY-003 — API Token Management & Scope Access Control
+
+- **Domain**: `security`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Generates scoped API bearer tokens (read/write/admin) for external tool and script authentication.
+
+### Evidence summary
+
+- `routes/api_token_routes.py` — `setup_api_token_routes` — Exposes API token creation, scope assignment, and revocation.
+- `core/database.py` — `ApiToken` — SQLAlchemy ORM schema for API tokens and permissions.
+
+### Unknowns
+
+- Leaked API bearer tokens with excessive permission scopes.
+
+## SECURITY-004 — Prompt Security & Injection Defense Engine
+
+- **Domain**: `security`
+- **Status**: `verified`
+- **Evidence Maturity**: `E1`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Scans system prompts and external inputs for prompt injection attempts, jailbreaks, and sensitive data leaks.
+
+### Evidence summary
+
+- `src/prompt_security.py` — `untrusted_context_message` — Wraps untrusted context with guard delimiters and sets metadata.trusted = False.
+- `src/tool_security.py` — `NON_ADMIN_BLOCKED_TOOLS` — Enforces tool execution safety for non-admin user roles.
+- `tests/test_skill_index_prompt_injection.py` — `test_skill_index` — Verifies skill index descriptions cannot leak into trusted system prompts.
+- `tests/test_tool_output_prompt_injection.py` — `test_tool_output` — Tool output injection guards.
+
+### Unknowns
+
+- False positives blocking legitimate complex coding or security prompts.
+
+## SECURITY-005 — URL & Path Confinement Security Guards
+
+- **Domain**: `security`
+- **Status**: `verified`
+- **Evidence Maturity**: `E1`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Prevents SSRF attacks and path traversal by validating target IP addresses and resolving symlinks.
+
+### Evidence summary
+
+- `src/url_safety.py` — `check_outbound_url` — Rejects non-HTTP(S) schemes, link-local, cloud metadata SSRF addresses.
+- `src/url_security.py` — `validate_public_http_url` — Validates public-facing endpoints.
+- `tests/test_url_safety.py` — `test_url_safety` — Scheme validation, cloud metadata SSRF rejection, IP classification.
+- `tests/test_tool_path_confinement.py` — `test_path_confinement` — Path traversal checks.
+- `tests/test_workspace_confine.py` — `test_workspace_confine` — Workspace confinement checks.
+
+### Unknowns
+
+- DNS rebinding attacks if IP address is re-resolved post-validation.
+
+## SECURITY-006 — HTTP Security Headers Middleware
+
+- **Domain**: `security`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Injects standard OWASP HTTP security headers (CSP, HSTS, X-Content-Type-Options, X-Frame-Options).
+
+### Evidence summary
+
+- `core/middleware.py` — `SecurityHeadersMiddleware` — Sets strict security headers and CSP nonces on HTTP responses.
+
+### Unknowns
+
+- Strict Content Security Policy (CSP) blocking third-party embedded web resources.
+
+## SECURITY-007 — Admin System Data Wipe ('Danger Zone')
+
+- **Domain**: `security`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Provides administrative reset operations to wipe sessions, cache, uploaded files, or factory reset state.
+
+### Evidence summary
+
+- `routes/admin_wipe/admin_wipe_routes.py` — `@router.delete('/wipe/{kind}')` — Executes systemic data wipe based on requested scope.
+
+### Unknowns
+
+- Accidental catastrophic data loss if triggered without user confirmation.

docs/discovery/domains/skill.md

@@ -0,0 +1,24 @@
+# Skill
+
+Features in this document are generated from [`../feature-catalog.json`](../feature-catalog.json), the canonical inventory.
+
+## SKILL-001 — Dynamic Skill Management & Code Execution Engine
+
+- **Domain**: `skill`
+- **Status**: `verified`
+- **Evidence Maturity**: `E0`
+- **Commit Verified**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Runtime Validation**: not-required — No separate environment-dependent runtime validation was identified during this documentation pass.
+
+### Purpose
+
+Allows users to create, import, edit, test, and execute custom Python/Markdown skills dynamically.
+
+### Evidence summary
+
+- `routes/skills_routes.py` — `setup_skills_routes` — Exposes CRUD and remote import routes for user skills.
+- `services/memory/skills.py` — `SkillsManager` — Handles skill storage, parsing, and execution.
+
+### Unknowns
+
+- Arbitrary code execution risks if skill import URL is untrusted.

docs/discovery/feature-catalog.md

@@ -0,0 +1,85 @@
+# Feature Catalog
+
+This is a human-readable index derived from [`feature-catalog.json`](feature-catalog.json). The JSON file is canonical.
+
+| ID | Feature | Domain | Status | Evidence | Runtime |
+|---|---|---|---|---|---|
+| `AGENT-001` | Autonomous Agent Loop & Tool Execution Engine | `agent` | `verified` | `E0` | not required |
+| `AGENT-002` | Scheduled Tasks & Event Bus Dispatcher | `agent` | `verified` | `E0` | not required |
+| `AGENT-003` | Webhook Event Subscriptions & Trigger Processing | `agent` | `verified` | `E0` | not required |
+| `AGENT-004` | Assistant Settings, Task Check-Ins & Background Job Monitor | `agent` | `partial` | `E1` | not required |
+| `AGENT-005` | Model Context Protocol (MCP) Server Integration | `agent` | `verified` | `E0` | not required |
+| `AGENT-006` | AI Interaction Tools & Pipeline Orchestration | `agent` | `verified` | `E0` | not required |
+| `AGENT-007` | Subprocess & Background Job Execution Tools | `agent` | `verified` | `E0` | not required |
+| `CALENDAR-001` | CalDAV Calendar Synchronization & Account Setup | `calendar` | `verified` | `E0` | pending |
+| `CALENDAR-002` | Calendar Event Operations & iCalendar Parsing | `calendar` | `verified` | `E0` | not required |
+| `CHAT-001` | Core Chat Streaming & SSE Message Generation | `chat` | `verified` | `E2` | pending |
+| `CHAT-002` | Session Management & Conversation State | `chat` | `verified` | `E0` | not required |
+| `CHAT-003` | Chat History & Message Editing/Truncation | `chat` | `verified` | `E0` | not required |
+| `CHAT-004` | File & Multimodal Attachment Handling | `chat` | `verified` | `E0` | not required |
+| `CHAT-005` | Chat Message Search | `chat` | `verified` | `E0` | not required |
+| `CHAT-006` | System Prompts & Preset Management | `chat` | `verified` | `E0` | not required |
+| `CHAT-007` | Emoji Rendering & Twemoji SVG Proxy | `chat` | `verified` | `E0` | not required |
+| `CHAT-008` | Input History Recall (Arrow Up) | `chat` | `verified` | `E0` | not required |
+| `CHAT-009` | Context Window Compaction & Truncation | `chat` | `verified` | `E0` | not required |
+| `CONTACT-001` | CardDAV Contact Management & Address Book Integration | `contact` | `verified` | `E0` | not required |
+| `COOKBOOK-001` | Local Model Download & Recipe Lifecycle Management | `cookbook` | `verified` | `E0` | not required |
+| `COOKBOOK-002` | Hardware Model Fitting ('What Fits?') Analysis Engine | `cookbook` | `verified` | `E0` | not required |
+| `COOKBOOK-003` | HuggingFace & MLX Model Discovery Services | `cookbook` | `verified` | `E0` | not required |
+| `COOKBOOK-004` | Host Docker Access for Inference Container Runtimes | `cookbook` | `verified` | `E0` | pending |
+| `DOCUMENT-001` | Document & Canvas Artifact Management | `document` | `verified` | `E0` | not required |
+| `DOCUMENT-002` | PDF Form Processing & High-Fidelity Rendering | `document` | `verified` | `E1` | pending |
+| `DOCUMENT-003` | Personal Document Indexing & RAG Retrieval | `document` | `verified` | `E0` | not required |
+| `DOCUMENT-004` | Document Conversion & Text Extraction Engine | `document` | `verified` | `E0` | not required |
+| `DOCUMENT-005` | Document Library UI Navigation | `document` | `verified` | `E0` | not required |
+| `EMAIL-001` | Email Account Setup, IMAP/SMTP Connection & Polling | `email` | `verified` | `E1` | pending |
+| `EMAIL-002` | Email Searching, Threading & Message Operations | `email` | `verified` | `E0` | not required |
+| `EMAIL-003` | Email Composition, Draft Management & Sending | `email` | `verified` | `E0` | pending |
+| `EMAIL-004` | Email MCP Server & Codex Integration Bridge | `email` | `verified` | `E0` | not required |
+| `FRONTEND-001` | Single Page Application Shell & Client Router | `frontend` | `verified` | `E0` | not required |
+| `FRONTEND-002` | Dynamic Theme, Color System & Custom Fonts | `frontend` | `verified` | `E0` | not required |
+| `FRONTEND-003` | Window Manager, Tile Layout & Modal Control System | `frontend` | `verified` | `E0` | not required |
+| `FRONTEND-004` | Global Keyboard Shortcuts & Accessibility Controls | `frontend` | `verified` | `E0` | not required |
+| `FRONTEND-005` | Markdown, LaTeX & Code Block Streaming Renderer | `frontend` | `verified` | `E0` | not required |
+| `FRONTEND-006` | Interactive Tour & Guided Onboarding System | `frontend` | `verified` | `E0` | not required |
+| `FRONTEND-007` | Background Effects Prototyping Sandbox | `frontend` | `dead-code-candidate` | `E1` | not required |
+| `MEDIA-001` | Gallery Image Library & Album Operations | `media` | `verified` | `E0` | not required |
+| `MEDIA-002` | Image Processing, AI Upscaling & Style Transfer | `media` | `verified` | `E0` | not required |
+| `MEDIA-003` | Interactive Image Canvas Editor & Persisted Drafts | `media` | `verified` | `E0` | not required |
+| `MEDIA-004` | Text-to-Speech (TTS) Synthesis Service | `media` | `verified` | `E0` | not required |
+| `MEDIA-005` | Speech-to-Text (STT) Audio Transcription Service | `media` | `verified` | `E0` | not required |
+| `MEDIA-006` | Digital Signature Stamp Storage & Placement | `media` | `verified` | `E0` | not required |
+| `MEDIA-007` | Generated Image Artifact Route & MCP Integration | `media` | `verified` | `E0` | not required |
+| `MEDIA-008` | Native MLX Image Bridge (macOS Apple Silicon) | `media` | `experimental` | `E0` | pending |
+| `MEMORY-001` | Persistent Long-Term Memory & Vector Indexing | `memory` | `verified` | `E0` | not required |
+| `MODEL-001` | Multi-Provider LLM Model Discovery & Metadata Management | `model` | `verified` | `E0` | not required |
+| `MODEL-002` | Model Capability & Context Limits Detection | `model` | `verified` | `E0` | not required |
+| `MODEL-003` | LLM Core Provider Communication & Fallback Routing | `model` | `verified` | `E0` | pending |
+| `MODEL-004` | Model Selection & Display Ordering Preferences | `model` | `verified` | `E0` | not required |
+| `MODEL-005` | Side-by-Side Model Comparison (A/B Testing) | `model` | `verified` | `E0` | not required |
+| `MODEL-006` | GitHub Copilot Device Flow Authentication | `model` | `verified` | `E2` | pending |
+| `MODEL-007` | ChatGPT Subscription Device Flow Authentication | `model` | `verified` | `E2` | pending |
+| `MODEL-008` | Embedding Model Lane & Vector Provider Setup | `model` | `verified` | `E0` | not required |
+| `NOTE-001` | Interactive Notes & Checklist Management | `note` | `verified` | `E0` | not required |
+| `PLATFORM-001` | Application Initialization & Lifespan Management | `platform` | `verified` | `E0` | not required |
+| `PLATFORM-002` | System Health, Readiness & Version Monitoring APIs | `platform` | `verified` | `E0` | not required |
+| `PLATFORM-003` | Database Schema, Migrations & SQLite Persistence | `platform` | `verified` | `E0` | not required |
+| `PLATFORM-004` | User Data Export & Import Backup Infrastructure | `platform` | `verified` | `E0` | not required |
+| `PLATFORM-005` | File Cleanup & Storage Maintenance Engine | `platform` | `verified` | `E0` | not required |
+| `PLATFORM-006` | System Health & RAG Diagnostic Suite | `platform` | `verified` | `E0` | not required |
+| `PLATFORM-007` | Desktop CLI Utilities & Shell Integration Tools | `platform` | `verified` | `E0` | not required |
+| `PLATFORM-008` | Desktop Companion App Integration | `platform` | `verified` | `E0` | not required |
+| `PLATFORM-009` | Docker Containerization & GPU Hardware Manifests | `platform` | `verified` | `E0` | pending |
+| `PLATFORM-010` | Legacy FAISS Vector Index Migration Script | `platform` | `legacy` | `E0` | not required |
+| `RESEARCH-001` | Deep Research Execution Engine & SSE Progress Streaming | `research` | `verified` | `E0` | not required |
+| `RESEARCH-002` | Research Library, Detail View & Image Controls | `research` | `verified` | `E0` | not required |
+| `RESEARCH-003` | Web Search Engine Integration (SearXNG & Multi-Provider) | `research` | `verified` | `E1` | pending |
+| `RESEARCH-004` | Research Result Peeking & Topic Spinoff Generation | `research` | `verified` | `E0` | not required |
+| `SECURITY-001` | Authentication, Session Cookies & User Management | `security` | `verified` | `E0` | not required |
+| `SECURITY-002` | System Vault Encrypted Secret Storage | `security` | `verified` | `E1` | pending |
+| `SECURITY-003` | API Token Management & Scope Access Control | `security` | `verified` | `E0` | not required |
+| `SECURITY-004` | Prompt Security & Injection Defense Engine | `security` | `verified` | `E1` | not required |
+| `SECURITY-005` | URL & Path Confinement Security Guards | `security` | `verified` | `E1` | not required |
+| `SECURITY-006` | HTTP Security Headers Middleware | `security` | `verified` | `E0` | not required |
+| `SECURITY-007` | Admin System Data Wipe ('Danger Zone') | `security` | `verified` | `E0` | not required |
+| `SKILL-001` | Dynamic Skill Management & Code Execution Engine | `skill` | `verified` | `E0` | not required |

docs/discovery/references/source-provenance.md

@@ -0,0 +1,28 @@
+# Source Provenance & Audit Baseline
+
+## Target Repository & Snapshot
+
+- **Repository**: `odysseus-dev/odysseus`
+- **Branch**: `discovery`
+- **Pinned Commit SHA**: `d8a2059df8e53bc7275c45339849d14c8651e73c`
+- **Snapshot Date**: `2026-07-23T14:49:02Z`
+
+## Discovery Package Organization
+
+The public discovery documentation package under `docs/discovery/` is structured as follows:
+
+- `feature-catalog.json`: Canonical machine-readable JSON catalog containing 79 feature records.
+- `feature-catalog.md`: Human-readable summary derived from `feature-catalog.json`.
+- `BASELINE-STATUS.md`: Publication status, evidence-validation snapshot, and durable maintainer guidance.
+- `audit-method.md`: Audit rules, scope, and evidence maturity definitions (E0 to E4).
+- `domains/`: 16 functional domain markdown files detailing feature implementations.
+- `references/`: Audit provenance and repository snapshot metadata.
+- `tools/`: Structural, consistency, and evidence validators with focused evidence-validator tests.
+
+## Exclusion Principles
+
+This public documentation package explicitly excludes:
+- Internal planning artifacts or private meeting notes.
+- Machine-specific filesystem paths or user environments.
+- API keys, credentials, or private service endpoints.
+- Application code or automated test mutations.

docs/discovery/tools/test_validate_discovery_evidence.py

@@ -0,0 +1,224 @@
+#!/usr/bin/env python3
+"""Focused negative tests for validate_discovery_evidence.py."""
+
+from __future__ import annotations
+
+import importlib.util
+import shutil
+import sys
+import tempfile
+import unittest
+from pathlib import Path
+
+
+MODULE_PATH = Path(__file__).with_name("validate_discovery_evidence.py")
+SPEC = importlib.util.spec_from_file_location("validate_discovery_evidence", MODULE_PATH)
+assert SPEC and SPEC.loader
+validator = importlib.util.module_from_spec(SPEC)
+sys.modules[SPEC.name] = validator
+SPEC.loader.exec_module(validator)
+
+
+class FakeJavascriptParser:
+    supported = True
+    reason = "test parser"
+
+    def __init__(self, symbols: list[validator.Located] | None = None) -> None:
+        self.symbols = symbols or []
+
+    def parse(self, path: Path) -> list[validator.Located]:
+        return self.symbols
+
+
+class UnsupportedJavascriptParser:
+    supported = False
+    reason = "no repository-local parser"
+
+
+class EvidenceNegativeTests(unittest.TestCase):
+    def setUp(self) -> None:
+        self.temp = tempfile.TemporaryDirectory()
+        self.root = Path(self.temp.name)
+        (self.root / "sample.py").write_text(
+            "from fastapi import APIRouter\n"
+            "router = APIRouter(prefix='/api')\n"
+            "\n"
+            "class ChatHandler:\n"
+            "    def preprocess_message(self):\n"
+            "        return True\n"
+            "\n"
+            "@router.post('/chat')\n"
+            "def chat_stream():\n"
+            "    return True\n",
+            encoding="utf-8",
+        )
+        (self.root / "sample.js").write_text(
+            "export const present = () => true;\n", encoding="utf-8"
+        )
+        (self.root / "sample.sh").write_text(
+            "#!/usr/bin/env bash\nreal_function() {\n  return 0\n}\n",
+            encoding="utf-8",
+        )
+        self.backup = self.root / "copied-backups"
+        self.backup.mkdir()
+        for path in self.root.glob("sample.*"):
+            shutil.copy2(path, self.backup / path.name)
+
+    def tearDown(self) -> None:
+        for backup in self.backup.iterdir():
+            target = self.root / backup.name
+            shutil.copy2(backup, target)
+            self.assertEqual(target.read_bytes(), backup.read_bytes())
+        self.temp.cleanup()
+
+    def validate(
+        self,
+        evidence: dict[str, str],
+        javascript_parser: object | None = None,
+    ) -> validator.Validation:
+        return validator.validate_evidence(
+            self.root,
+            "TEST-001",
+            0,
+            evidence,
+            javascript_parser or FakeJavascriptParser(),
+        )
+
+    def test_missing_python_symbol(self) -> None:
+        result = self.validate(
+            {
+                "path": "sample.py",
+                "kind": "python-function",
+                "locator": "fabricated",
+                "line_range": "L1-L1",
+                "explanation": "negative fixture",
+            }
+        )
+        self.assertEqual(result.result, "invalid-locator")
+
+    def test_incorrect_qualified_method(self) -> None:
+        result = self.validate(
+            {
+                "path": "sample.py",
+                "kind": "python-method",
+                "locator": "WrongHandler.preprocess_message",
+                "line_range": "L5-L6",
+                "explanation": "negative fixture",
+            }
+        )
+        self.assertEqual(result.result, "invalid-locator")
+
+    def test_symbol_outside_cited_range(self) -> None:
+        result = self.validate(
+            {
+                "path": "sample.py",
+                "kind": "python-method",
+                "locator": "ChatHandler.preprocess_message",
+                "line_range": "L1-L2",
+                "explanation": "negative fixture",
+            }
+        )
+        self.assertEqual(result.result, "locator-outside-range")
+
+    def test_fabricated_test_function(self) -> None:
+        result = self.validate(
+            {
+                "path": "sample.py",
+                "kind": "test-function",
+                "locator": "test_fabricated",
+                "line_range": "L1-L2",
+                "explanation": "negative fixture",
+            }
+        )
+        self.assertEqual(result.result, "invalid-locator")
+
+    def test_nonexistent_javascript_symbol_with_parser(self) -> None:
+        result = self.validate(
+            {
+                "path": "sample.js",
+                "kind": "javascript-function",
+                "locator": "missing",
+                "line_range": "L1-L1",
+                "explanation": "negative fixture",
+            },
+            FakeJavascriptParser(
+                [validator.Located("present", "javascript-export", 1, 1)]
+            ),
+        )
+        self.assertEqual(result.result, "invalid-locator")
+
+    def test_unsupported_javascript_parser(self) -> None:
+        result = self.validate(
+            {
+                "path": "sample.js",
+                "kind": "javascript-function",
+                "locator": "present",
+                "line_range": "L1-L1",
+                "explanation": "negative fixture",
+            },
+            UnsupportedJavascriptParser(),
+        )
+        self.assertEqual(result.result, "unsupported")
+
+    def test_route_path_mismatch(self) -> None:
+        result = self.validate(
+            {
+                "path": "sample.py",
+                "kind": "python-route",
+                "locator": "POST /api/wrong -> chat_stream",
+                "line_range": "L9-L10",
+                "explanation": "negative fixture",
+            }
+        )
+        self.assertEqual(result.result, "invalid-locator")
+        self.assertIn("path", result.problem or "")
+
+    def test_http_method_mismatch(self) -> None:
+        result = self.validate(
+            {
+                "path": "sample.py",
+                "kind": "python-route",
+                "locator": "GET /api/chat -> chat_stream",
+                "line_range": "L9-L10",
+                "explanation": "negative fixture",
+            }
+        )
+        self.assertEqual(result.result, "invalid-locator")
+        self.assertIn("method", result.problem or "")
+
+    def test_shell_function_mismatch(self) -> None:
+        result = self.validate(
+            {
+                "path": "sample.sh",
+                "kind": "shell-function",
+                "locator": "fabricated",
+                "line_range": "L1-L4",
+                "explanation": "negative fixture",
+            }
+        )
+        self.assertEqual(result.result, "invalid-locator")
+
+    def test_invalid_file_level_evidence(self) -> None:
+        result = self.validate(
+            {
+                "path": "missing.file",
+                "kind": "file",
+                "explanation": "negative fixture",
+            }
+        )
+        self.assertEqual(result.result, "invalid-path")
+
+    def test_file_level_evidence_rejects_fake_symbol(self) -> None:
+        result = self.validate(
+            {
+                "path": "sample.sh",
+                "kind": "file",
+                "locator": "whole-script",
+                "explanation": "negative fixture",
+            }
+        )
+        self.assertEqual(result.result, "invalid-locator")
+
+
+if __name__ == "__main__":
+    unittest.main()

docs/discovery/tools/validate_discovery_consistency.py

@@ -0,0 +1,592 @@
+#!/usr/bin/env python3
+
+from __future__ import annotations
+
+import json
+import re
+import sys
+from collections import Counter, defaultdict
+from pathlib import Path
+from typing import Any
+
+
+ROOT = Path(__file__).resolve().parents[1]
+CATALOG_PATH = ROOT / "feature-catalog.json"
+CATALOG_MD_PATH = ROOT / "feature-catalog.md"
+DOMAINS_DIR = ROOT / "domains"
+REVIEWS_DIR = ROOT / "reviews"
+
+EXPECTED_COMMIT = "d8a2059df8e53bc7275c45339849d14c8651e73c"
+EXPECTED_FEATURES = 79
+EXPECTED_DOMAINS = 16
+
+VALID_STATUSES = {
+    "verified",
+    "partial",
+    "disabled",
+    "experimental",
+    "legacy",
+    "dead-code-candidate",
+    "unverified",
+}
+
+VALID_MATURITY = {"E0", "E1", "E2", "E3", "E4"}
+
+VALID_RUNTIME = {
+    "not-required",
+    "pending",
+    "blocked",
+    "completed",
+}
+
+REQUIRED_FIELDS = {
+    "id",
+    "domain",
+    "name",
+    "purpose",
+    "status",
+    "evidence_maturity",
+    "verified_at_commit",
+    "evidence",
+    "runtime_validation",
+}
+
+FEATURE_ID_RE = re.compile(r"^[A-Z][A-Z0-9]*-\d{3}$")
+
+DOMAIN_HEADING_RE = re.compile(
+    r"^##\s+`?([A-Z][A-Z0-9]*-\d{3})`?"
+    r"\s+(?:—|-)\s+(.+?)\s*$"
+)
+
+DOMAIN_FIELD_RE = re.compile(
+    r"^-\s+\*\*"
+    r"(Domain|Status|Evidence Maturity|Commit Verified)"
+    r"\*\*:\s*(.*?)\s*$"
+)
+
+DOMAIN_RUNTIME_RE = re.compile(
+    r"^-\s+\*\*Runtime Validation\*\*:\s*(.*?)\s*$"
+)
+
+REVIEW_HEADING_RE = re.compile(
+    r"^###\s+(?:\d+\.\s+)?"
+    r"([A-Z][A-Z0-9]*-\d{3})"
+    r"\s+(?:—|-)\s+.+$"
+)
+
+REVIEW_RESULT_RE = re.compile(
+    r"^-\s+\*\*Resulting Status(?:\s+and|/)\s+Maturity\*\*:"
+    r"\s*`([^`]+)`\s*/\s*`([^`]+)`\s*$"
+)
+
+
+def clean(value: str) -> str:
+    value = value.strip()
+
+    if (
+        len(value) >= 2
+        and value.startswith("`")
+        and value.endswith("`")
+    ):
+        return value[1:-1].strip()
+
+    return value
+
+
+def load_catalog(errors: list[str]) -> list[dict[str, Any]]:
+    try:
+        data = json.loads(
+            CATALOG_PATH.read_text(encoding="utf-8")
+        )
+    except Exception as exc:
+        errors.append(f"Unable to read catalog: {exc}")
+        return []
+
+    if isinstance(data, list):
+        features = data
+    elif isinstance(data, dict) and isinstance(data.get("features"), list):
+        features = data["features"]
+    else:
+        errors.append(
+            "Catalog must be an array or contain a features array"
+        )
+        return []
+
+    if not all(isinstance(feature, dict) for feature in features):
+        errors.append("Every catalog feature must be an object")
+        return []
+
+    return features
+
+
+def validate_catalog(
+    features: list[dict[str, Any]],
+    errors: list[str],
+) -> None:
+    if len(features) != EXPECTED_FEATURES:
+        errors.append(
+            f"Expected {EXPECTED_FEATURES} features, "
+            f"found {len(features)}"
+        )
+
+    ids = [feature.get("id") for feature in features]
+
+    duplicates = sorted(
+        feature_id
+        for feature_id, count in Counter(ids).items()
+        if feature_id and count > 1
+    )
+
+    if duplicates:
+        errors.append(
+            "Duplicate feature IDs: " + ", ".join(duplicates)
+        )
+
+    for index, feature in enumerate(features):
+        feature_id = feature.get("id")
+        label = (
+            feature_id
+            if isinstance(feature_id, str)
+            else f"<index:{index}>"
+        )
+
+        missing = sorted(
+            field
+            for field in REQUIRED_FIELDS
+            if feature.get(field) in (None, "", [])
+        )
+
+        if missing:
+            errors.append(
+                f"{label}: missing fields: {', '.join(missing)}"
+            )
+
+        if (
+            not isinstance(feature_id, str)
+            or not FEATURE_ID_RE.fullmatch(feature_id)
+        ):
+            errors.append(f"{label}: invalid feature ID")
+
+        if feature.get("status") not in VALID_STATUSES:
+            errors.append(
+                f"{label}: invalid status "
+                f"{feature.get('status')!r}"
+            )
+
+        if feature.get("evidence_maturity") not in VALID_MATURITY:
+            errors.append(
+                f"{label}: invalid maturity "
+                f"{feature.get('evidence_maturity')!r}"
+            )
+
+        if feature.get("verified_at_commit") != EXPECTED_COMMIT:
+            errors.append(
+                f"{label}: incorrect verified_at_commit"
+            )
+
+        runtime = feature.get("runtime_validation")
+
+        if not isinstance(runtime, dict):
+            errors.append(
+                f"{label}: runtime_validation must be an object"
+            )
+            continue
+
+        required = runtime.get("required")
+        runtime_status = runtime.get("status")
+        reason = runtime.get("reason")
+
+        if not isinstance(required, bool):
+            errors.append(
+                f"{label}: runtime required must be boolean"
+            )
+
+        if runtime_status not in VALID_RUNTIME:
+            errors.append(
+                f"{label}: invalid runtime status "
+                f"{runtime_status!r}"
+            )
+
+        if not isinstance(reason, str) or not reason.strip():
+            errors.append(
+                f"{label}: runtime reason is blank"
+            )
+
+        if required is False and runtime_status != "not-required":
+            errors.append(
+                f"{label}: required=false requires not-required"
+            )
+
+        if required is True and runtime_status == "not-required":
+            errors.append(
+                f"{label}: required=true cannot be not-required"
+            )
+
+
+def parse_domain(
+    path: Path,
+    errors: list[str],
+) -> dict[str, dict[str, str]]:
+    records: dict[str, dict[str, str]] = {}
+    current_id: str | None = None
+
+    for line_number, line in enumerate(
+        path.read_text(encoding="utf-8").splitlines(),
+        start=1,
+    ):
+        heading = DOMAIN_HEADING_RE.match(line)
+
+        if heading:
+            current_id = heading.group(1)
+
+            if current_id in records:
+                errors.append(
+                    f"{path.relative_to(ROOT)}:{line_number}: "
+                    f"duplicate heading {current_id}"
+                )
+
+            records[current_id] = {
+                "Name": heading.group(2).strip(),
+            }
+            continue
+
+        field = DOMAIN_FIELD_RE.match(line)
+
+        if field and current_id:
+            value = clean(field.group(2))
+
+            if not value:
+                errors.append(
+                    f"{path.relative_to(ROOT)}:{line_number}: "
+                    f"blank {field.group(1)}"
+                )
+
+            records[current_id][field.group(1)] = value
+            continue
+
+        runtime = DOMAIN_RUNTIME_RE.match(line)
+
+        if runtime and current_id:
+            value = clean(runtime.group(1))
+            runtime_status = re.split(
+                r"\s+(?:—|-)\s+",
+                value,
+                maxsplit=1,
+            )[0]
+            records[current_id]["Runtime Validation"] = (
+                runtime_status.strip("` ")
+            )
+
+    return records
+
+
+def validate_domains(
+    features: list[dict[str, Any]],
+    errors: list[str],
+) -> None:
+    catalog = {
+        feature["id"]: feature
+        for feature in features
+        if feature.get("id")
+    }
+
+    expected_by_domain: defaultdict[str, set[str]] = defaultdict(set)
+
+    for feature in features:
+        expected_by_domain[feature["domain"]].add(feature["id"])
+
+    paths = sorted(DOMAINS_DIR.glob("*.md"))
+
+    if len(paths) != EXPECTED_DOMAINS:
+        errors.append(
+            f"Expected {EXPECTED_DOMAINS} domain files, "
+            f"found {len(paths)}"
+        )
+
+    all_found: set[str] = set()
+
+    for path in paths:
+        domain = path.stem
+        records = parse_domain(path, errors)
+        found = set(records)
+        expected = expected_by_domain.get(domain, set())
+        all_found.update(found)
+
+        if found != expected:
+            missing = sorted(expected - found)
+            extra = sorted(found - expected)
+
+            errors.append(
+                f"{domain}: missing={missing}, unexpected={extra}"
+            )
+
+        for feature_id, record in records.items():
+            feature = catalog.get(feature_id)
+
+            if feature is None:
+                continue
+
+            expected_values = {
+                "Name": feature["name"],
+                "Domain": feature["domain"],
+                "Status": feature["status"],
+                "Evidence Maturity": feature["evidence_maturity"],
+                "Commit Verified": feature["verified_at_commit"],
+                "Runtime Validation": (
+                    feature["runtime_validation"]["status"]
+                ),
+            }
+
+            for field, expected_value in expected_values.items():
+                actual = record.get(field)
+
+                if actual != expected_value:
+                    errors.append(
+                        f"{path.relative_to(ROOT)}: "
+                        f"{feature_id} {field}: "
+                        f"{actual!r} != {expected_value!r}"
+                    )
+
+    if all_found != set(catalog):
+        errors.append(
+            "Domain Markdown IDs do not match catalog JSON"
+        )
+
+
+def validate_catalog_markdown(
+    features: list[dict[str, Any]],
+    errors: list[str],
+) -> None:
+    expected = {
+        feature["id"]: feature
+        for feature in features
+    }
+    found: dict[str, list[str]] = {}
+
+    for line in CATALOG_MD_PATH.read_text(
+        encoding="utf-8"
+    ).splitlines():
+        if not line.startswith("|"):
+            continue
+
+        cells = [
+            cell.strip()
+            for cell in line.strip().strip("|").split("|")
+        ]
+
+        if len(cells) < 6:
+            continue
+
+        feature_id = clean(cells[0])
+
+        if FEATURE_ID_RE.fullmatch(feature_id):
+            found[feature_id] = cells
+
+    if set(found) != set(expected):
+        errors.append(
+            "feature-catalog.md IDs do not match JSON"
+        )
+
+    for feature_id, cells in found.items():
+        feature = expected[feature_id]
+        runtime = feature["runtime_validation"]
+        runtime_display = (
+            runtime["status"]
+            if runtime["required"]
+            else "not required"
+        )
+
+        actual = {
+            "name": cells[1].replace("\\|", "|"),
+            "domain": clean(cells[2]),
+            "status": clean(cells[3]),
+            "maturity": clean(cells[4]),
+            "runtime": clean(cells[5]),
+        }
+
+        wanted = {
+            "name": feature["name"],
+            "domain": feature["domain"],
+            "status": feature["status"],
+            "maturity": feature["evidence_maturity"],
+            "runtime": runtime_display,
+        }
+
+        for field, expected_value in wanted.items():
+            if actual[field] != expected_value:
+                errors.append(
+                    f"feature-catalog.md: {feature_id} "
+                    f"{field}: {actual[field]!r} "
+                    f"!= {expected_value!r}"
+                )
+
+
+def validate_reviews(
+    features: list[dict[str, Any]],
+    errors: list[str],
+) -> int:
+    catalog = {
+        feature["id"]: feature
+        for feature in features
+    }
+    checked = 0
+
+    for path in sorted(
+        REVIEWS_DIR.glob("evidence-sample-*.md")
+    ):
+        current_id: str | None = None
+        results: set[str] = set()
+
+        for line_number, line in enumerate(
+            path.read_text(encoding="utf-8").splitlines(),
+            start=1,
+        ):
+            heading = REVIEW_HEADING_RE.match(line)
+
+            if heading:
+                current_id = heading.group(1)
+                continue
+
+            result = REVIEW_RESULT_RE.match(line)
+
+            if not result or current_id is None:
+                continue
+
+            status, maturity = result.groups()
+            results.add(current_id)
+            checked += 1
+
+            feature = catalog.get(current_id)
+
+            if feature is None:
+                errors.append(
+                    f"{path.relative_to(ROOT)}:{line_number}: "
+                    f"unknown feature {current_id}"
+                )
+                continue
+
+            if feature["status"] != status:
+                errors.append(
+                    f"{current_id}: review status {status!r} "
+                    f"!= catalog {feature['status']!r}"
+                )
+
+            if feature["evidence_maturity"] != maturity:
+                errors.append(
+                    f"{current_id}: review maturity {maturity!r} "
+                    f"!= catalog "
+                    f"{feature['evidence_maturity']!r}"
+                )
+
+        if path.name == "evidence-sample-01.md" and len(results) != 12:
+            errors.append(
+                f"{path.relative_to(ROOT)}: expected 12 "
+                f"review results, found {len(results)}"
+            )
+
+    return checked
+
+
+def validate_whitespace(errors: list[str]) -> None:
+    for path in sorted(ROOT.rglob("*")):
+        if not path.is_file():
+            continue
+
+        if path.suffix not in {".md", ".json", ".py", ".txt"}:
+            continue
+
+        for line_number, line in enumerate(
+            path.read_text(
+                encoding="utf-8",
+                errors="replace",
+            ).splitlines(),
+            start=1,
+        ):
+            if line != line.rstrip(" \t"):
+                errors.append(
+                    f"{path.relative_to(ROOT)}:{line_number}: "
+                    "trailing whitespace"
+                )
+
+    if any(ROOT.rglob("*.pyc")):
+        errors.append("Generated .pyc files exist")
+
+    if any(
+        path.is_dir()
+        for path in ROOT.rglob("__pycache__")
+    ):
+        errors.append("__pycache__ exists")
+
+
+def main() -> int:
+    errors: list[str] = []
+    features = load_catalog(errors)
+    reviewed = 0
+
+    if features:
+        validate_catalog(features, errors)
+        validate_domains(features, errors)
+        validate_catalog_markdown(features, errors)
+        reviewed = validate_reviews(features, errors)
+
+    validate_whitespace(errors)
+
+    print(f"Catalog Features: {len(features)}")
+    print(
+        "Unique Feature IDs:",
+        len({feature.get("id") for feature in features}),
+    )
+    print(
+        "Domain Files:",
+        len(list(DOMAINS_DIR.glob("*.md"))),
+    )
+
+    if features:
+        print(
+            "Statuses:",
+            dict(
+                Counter(
+                    feature.get("status")
+                    for feature in features
+                )
+            ),
+        )
+        print(
+            "Evidence Maturity:",
+            dict(
+                Counter(
+                    feature.get("evidence_maturity")
+                    for feature in features
+                )
+            ),
+        )
+        print(
+            "Runtime Validation:",
+            dict(
+                Counter(
+                    (
+                        feature.get(
+                            "runtime_validation",
+                            {},
+                        ).get("required"),
+                        feature.get(
+                            "runtime_validation",
+                            {},
+                        ).get("status"),
+                    )
+                    for feature in features
+                )
+            ),
+        )
+
+    print(f"Review Results Checked: {reviewed}")
+    print(f"Consistency Errors: {len(errors)}")
+
+    for error in errors:
+        print(f"ERROR: {error}")
+
+    return 1 if errors else 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

docs/discovery/tools/validate_discovery_docs.py

@@ -0,0 +1,172 @@
+#!/usr/bin/env python3
+"""Validate the Odysseus public discovery documentation package."""
+
+from __future__ import annotations
+
+import argparse
+import json
+import re
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+REPO_ROOT = ROOT.parents[1]
+CATALOG_PATH = ROOT / "feature-catalog.json"
+DOMAINS_DIR = ROOT / "domains"
+
+EXPECTED_COMMIT = "d8a2059df8e53bc7275c45339849d14c8651e73c"
+ALLOWED_STATUSES = {
+    "verified", "partial", "disabled", "experimental", "legacy",
+    "dead-code-candidate", "unverified"
+}
+
+LINE_RANGE_RE = re.compile(r"^L([1-9]\d*)-L([1-9]\d*)$")
+FEATURE_HEADING_RE = re.compile(
+    r"^##\s+`?([A-Z][A-Z0-9]*-\d{3})`?\s+(?:—|-)\s+.+$",
+    re.MULTILINE,
+)
+LINK_RE = re.compile(r"(?<!!)\[[^\]]*\]\(([^)]+)\)")
+FORBIDDEN_TERMS_RE = re.compile(
+    r"(roadforge|kanban|matrix|owner link|editor link|github support|"
+    r"private maintainer|OD-AUD-|[A-Z]{2,10}-AUD-\d+|TASK-\d+)",
+    re.IGNORECASE,
+)
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description="Validate Odysseus discovery docs.")
+    parser.add_argument(
+        "--repo-root",
+        type=Path,
+        default=REPO_ROOT,
+        help="Path to Odysseus repository root.",
+    )
+    return parser.parse_args()
+
+def validate() -> int:
+    args = parse_args()
+    repo_root = args.repo_root.resolve()
+    errors: list[str] = []
+
+    # 1. Validate Catalog JSON existence and content
+    if not CATALOG_PATH.is_file():
+        errors.append(f"Missing catalog file: {CATALOG_PATH}")
+        print(f"Errors: {len(errors)}")
+        for e in errors:
+            print(f"ERROR: {e}")
+        return 1
+
+    try:
+        catalog = json.loads(CATALOG_PATH.read_text(encoding="utf-8"))
+    except Exception as e:
+        errors.append(f"Failed to parse catalog JSON: {e}")
+        print(f"Errors: {len(errors)}")
+        for err in errors:
+            print(f"ERROR: {err}")
+        return 1
+
+    if not isinstance(catalog, list):
+        errors.append("feature-catalog.json must be a JSON array")
+        return 1
+
+    feature_ids = [item.get("id") for item in catalog if isinstance(item, dict)]
+    if len(feature_ids) != 79:
+        errors.append(f"Expected 79 unique feature IDs, found {len(feature_ids)}")
+    if len(feature_ids) != len(set(feature_ids)):
+        errors.append("Duplicate feature IDs found in catalog JSON")
+
+    # Domain catalog counts
+    catalog_domain_counts: dict[str, int] = {}
+    for item in catalog:
+        if not isinstance(item, dict):
+            errors.append("Catalog item is not an object")
+            continue
+        fid = item.get("id", "<missing>")
+        status = item.get("status")
+        domain = item.get("domain", "").lower()
+        catalog_domain_counts[domain] = catalog_domain_counts.get(domain, 0) + 1
+
+        if status not in ALLOWED_STATUSES:
+            errors.append(f"{fid}: Invalid status '{status}'")
+
+        evidence_list = item.get("evidence")
+        if not isinstance(evidence_list, list) or not evidence_list:
+            errors.append(f"{fid}: Missing or empty evidence list")
+            continue
+
+        for ev in evidence_list:
+            if not isinstance(ev, dict):
+                errors.append(f"{fid}: Evidence item is not an object")
+                continue
+            path_str = ev.get("path")
+            lr_str = str(ev.get("line_range", ""))
+            if not path_str or Path(path_str).is_absolute() or ".." in Path(path_str).parts:
+                errors.append(f"{fid}: Unsafe or invalid path '{path_str}'")
+                continue
+
+            # Check path exists in repo
+            target_path = repo_root / path_str
+            if not target_path.is_file():
+                errors.append(f"{fid}: Referenced path '{path_str}' does not exist on disk")
+                continue
+
+            # Check line range format & bounds
+            m = LINE_RANGE_RE.fullmatch(lr_str)
+            if not m:
+                errors.append(f"{fid}: Invalid line range format '{lr_str}' for path '{path_str}'")
+                continue
+
+            start, end = int(m.group(1)), int(m.group(2))
+            lines_cnt = len(target_path.read_text(encoding="utf-8", errors="ignore").splitlines())
+            if start > end or end > lines_cnt or start < 1:
+                errors.append(
+                    f"{fid}: Line range '{lr_str}' exceeds file length ({lines_cnt} lines) in '{path_str}'"
+                )
+
+    # 2. Check Domain Markdown files
+    md_feature_ids: list[str] = []
+    domain_files = sorted(DOMAINS_DIR.glob("*.md"))
+    for df in domain_files:
+        domain_name = df.stem.lower()
+        content = df.read_text(encoding="utf-8")
+        found_ids = FEATURE_HEADING_RE.findall(content)
+        md_feature_ids.extend(found_ids)
+        if len(found_ids) != catalog_domain_counts.get(domain_name, 0):
+            errors.append(
+                f"Domain '{domain_name}' count mismatch: catalog has {catalog_domain_counts.get(domain_name, 0)}, Markdown has {len(found_ids)}"
+            )
+
+    if sorted(md_feature_ids) != sorted(feature_ids):
+        errors.append("Markdown domain feature IDs do not match catalog JSON feature IDs")
+
+    # 3. Check for forbidden/private terms, sensitive credentials, and broken links across all docs
+    for md_file in ROOT.rglob("*.md"):
+        rel_md = md_file.relative_to(ROOT)
+        content = md_file.read_text(encoding="utf-8")
+
+        # Forbidden terms scan
+        forbidden_matches = FORBIDDEN_TERMS_RE.findall(content)
+        if forbidden_matches:
+            errors.append(
+                f"{rel_md}: Found forbidden/internal terms: {set(forbidden_matches)}"
+            )
+
+        # Broken local link check
+        for target in LINK_RE.findall(content):
+            target = target.strip().strip("<>")
+            if not target or target.startswith(("#", "http://", "https://", "mailto:")):
+                continue
+            target_path = target.split("#", 1)[0]
+            resolved = (md_file.parent / target_path).resolve()
+            if not resolved.exists():
+                errors.append(f"{rel_md}: Broken local link '{target}'")
+
+    print(f"Catalog Features: {len(catalog)}")
+    print(f"Domain Files: {len(domain_files)}")
+    print(f"Validation Errors: {len(errors)}")
+    for err in errors:
+        print(f"ERROR: {err}")
+
+    return 1 if errors else 0
+
+if __name__ == "__main__":
+    sys.exit(validate())