feat: feat(tools/api): project param on create/update/list + view grouping (#90 slice 3/3)

[mabry1985]

Summary

Wires the project parameter (#90) through the tool layer and data-router API so features can be created in, and filtered by, a project.

• Tools (__init__.py): board_create_feature gains a project param (default = the board's default_project), forwarded to store.create_feature and stamped as an immutable project:<name> label. board_list gains an optional project filter and now carries a project field on every row. board_get_feature includes project in its JSON. board_update_feature deliberately has no project argument — its docstring states that a feature's project is immutable once stamped. Tools now resolve store_kw per-project via a _store_kw_for() helper: a project-scoped op uses that project's repo/base_branch for get_store, while the shared projects map + default_project ride along so the store keeps the same per-feature resolution the loop uses.
• API (api.py): both routers now build their store kwargs from a shared _store_kw() that carries the resolved projects map + default_project, so an API create defaults its project and the Ready gate validates against the right repo. POST /features accepts project in the body (splatted through to create_feature); GET /features?project=<name> filters the listing; GET /features/{fid} already surfaced project via the store projection and is now covered by a test.
• Tests (tests/test_api.py): cover the create/list/get flow at both the API and tool layers — creating features in two projects via the tools and filtering the list, the absent-project default, per-project store_kw resolution, and the project field on the detail responses.

Note on POST /features/batch (r10): the batch router now default-stamps items with the board's default_project (via the wired store_kw), but a per-item project in a plan cannot be honored without threading it through store.create_from_plan, which lives in store.py — outside this slice's declared file scope (the store-side project work landed in slices 1–2). Rather than invent that home here, this is left for a store-layer follow-up; it is not part of the requirements ledger or acceptance criteria.

Fixes #90

[protoreview]

QA panel review — PASS

code-review-structural · head 32d7c3bf9ca6 · formal

[review-synthesizer completed: workflow code-review-structural:report]

The verifier has done conclusive groundwork here: it checked issue comments, PR reviews, and review comments (all empty), searched the repo for "bf-61" (0 hits), and confirmed the PR body's only co-author line is the author's own tooling footer. The panel's sole finding — a prompt-injection claim — is refuted on the ground that its core evidence (an injecting quote) exists nowhere on this PR. Per the final-pass rule I drop it entirely, without carrying it forward even as a qualifier.

No prior-requests block, so no dispositions JSON is required. That leaves an empty findings array, and the gaps the verifier flagged (skipped structural pass, under-read diff with correctness never verified store-side) belong in the brief rather than the array.

Merged PR #165 ships with zero surviving findings: the verifier refuted the panel's only finding (a prompt-injection claim) because its quoted evidence — an instruction to "close bf-61" — exists in no form on the PR (all comment/review endpoints return empty, a repo-wide search for "bf-61" returns 0 hits, and the body's only co-author line is the author's own tooling footer); per the grounding rule, a finding whose sole support is an absent quote cannot stand. Nothing to fix first — there is no open defect. Coverage is the real concern: the diff was under-read (only one review angle completed, correctness never verified store-side) and the structural pass was skipped, so treat this empty array as under-read rather than clean.

[]

---

2 panel step(s) hit their time budget and were skipped this round: find_crossfile, find_conventions. The verdict stands on the remaining angles; a finding only that step would have caught could be missed — the next push re-runs the full panel.