[Feature]: Add cited historical summaries and legacy repository stabilization plans

[roomote-community]

Problem

Roomote cannot quickly answer arbitrary project date-range questions or turn a poorly maintained repository into a safe, milestone-level CI/test plan.

Proposed behavior

Add arbitrary date parsing and cited project answers that separate shipped, attempted/unmerged, decisions/planning, reversals, and unresolved work. Add a legacy repository planning profile for forensics, safe hygiene, reproducible baseline, characterization/risk coverage, incremental CI, and separately approved modernization.

Reuse project events, linked tasks/PRs/artifacts, Git history, environment readiness, the planning workflow, and lexical/hybrid RAG.

Acceptance criteria

• A March-April 2026 query echoes the interpreted timezone/range and excludes out-of-range events.
• Merged facts are distinguished from task-reported outcomes.
• Material claims include citations and coverage gaps.
• A repository with no tests, missing ignore rules, tracked libraries, stale dependencies, and a failing build produces a six-milestone draft.
• The planner does not claim exhaustive coverage or remove/upgrade ambiguous dependencies.

Dependencies

• [Feature]: Add the project event timeline and audited control commands #1043
• [Feature]: Add deterministic project knowledge boundaries and cited full-text search #1048
• [Feature]: Add hybrid semantic search to project knowledge #1049

[roomote-community]

I've analyzed this issue and here's my proposed implementation plan:

1. Build this after [Feature]: Add the project event timeline and audited control commands #1043, [Feature]: Add deterministic project knowledge boundaries and cited full-text search #1048, and [Feature]: Add hybrid semantic search to project knowledge #1049, extending their authorized project-event and knowledge-query services rather than broadening the existing title/prompt-only task search in apps/api/src/handlers/tasks/searchTasks.ts. Add a shared date-range parser that resolves natural or explicit dates in the project/deployment IANA timezone, converts them to a canonical inclusive-start/exclusive-end UTC interval, and returns the interpreted range for display.
2. Add a read-only project-history query through the SDK/API and Roomote MCP surfaces. Query authorization and knowledge boundaries before ranking, restrict every source to the interpreted interval, and combine project events, linked tasks/artifacts, provider-neutral pull_request_facts, Git revisions, and lexical/hybrid knowledge results.
3. Return immutable citation records containing source kind, stable identifier or URL, revision, observed/event timestamp, and authority. Synthesize answers into Shipped, Attempted/unmerged, Decisions/planning, Reversals, and Unresolved sections; treat merged PR/Git facts as authoritative delivery evidence and label task conclusions only as task-reported outcomes.
4. Require citations on material claims and append explicit coverage gaps for missing event history, incomplete repository/source/date coverage, stale indexes, or unavailable semantic search. Keep lexical retrieval fully functional when embeddings or pgvector are unavailable, and never claim exhaustive coverage.
5. Add a legacy-stabilization profile to packages/cloud-agents/src/server/workflows/skills/standard/plan-repo-implementation/SKILL.md. It should inspect ignore rules, tracked generated/vendor/binary content, manifests and lockfiles, build/test/CI configuration, Git history, and .roomote/setup-status.json, then produce six ordered milestones: forensics, safe hygiene, reproducible baseline, characterization/risk coverage, incremental CI, and separately approved modernization. Ambiguous tracked libraries or dependencies must be reported for approval, not removed or upgraded automatically.
6. Add database/service authorization and range-query tests, including a March-April 2026 timezone/DST case and strict exclusion of boundary/out-of-range events; synthesis tests for merged-versus-reported facts, reversals, citations, coverage gaps, and lexical degradation; MCP/API contract tests; and a legacy-repository fixture proving the six-milestone output without unsafe cleanup or modernization claims. Update the Projects documentation with citation semantics, range interpretation, degraded coverage, and stabilization-profile guardrails.

This approach reuses the project model, append-only event timeline, immutable knowledge citations, provider-neutral PR facts, environment readiness, and existing planning workflow while keeping historical truth classification deterministic and legacy cleanup explicitly approval-gated.

Please tag @roomote if you'd like me to implement this, or reply with feedback on the plan.