feat(agent): agent audit feed — surface executed actions + approval decisions (#784)

[oktofeesh1]

Closes #784. Second of two focused PRs for Phase 3's agent automation control surface (the first, #934, added the MCP approval-queue list/decide tools and is merged). Rebased onto current main; supersedes #935/#936 (closed) with the review feedback fully addressed.

What this adds

The agent already writes an audit trail (agent.action.<class>, agent.pending_action.accepted|rejected) but nothing could read it back. This adds the read side:

• DB: listAgentAuditEvents(repoFullName, since?, limit) — repo-scoped via the repo#<...> targetKey range, filtered to agent events, newest first, capped at 200.
• HTTP: GET /v1/repos/:owner/:repo/agent/audit-feed — maintainer-scoped, ?since=ISO&limit=N. Now in the OpenAPI spec.
• MCP: gittensory_get_agent_audit_feed {owner, repo, since?, limit?} — maintainer-manage scoped.

Review feedback addressed (from #936)

• Sanitisation: the free-form detail is scrubbed with sanitizePublicComment before it leaves the endpoint or the MCP tool (defense-in-depth on the public/private boundary). Other fields are controlled vocab (eventType/outcome) or logins (actor).
• since validation: HTTP returns 400 for a non-ISO-8601 since; the MCP schema validates it as z.string().datetime().
• limit bound: HTTP returns 400 for a non-integer or out-of-1–200 limit (no longer silently clamped); the MCP schema caps at 200.
• Robust prefix range: upper bound is repo# + U+FFFF so no delimiter-adjacent key is wrongly excluded.
• OpenAPI: the endpoint + its auth/validation responses are documented; npm run ui:openapi:check passes.

Scope on #784

With #934 (MCP list/decide) and the merged CLI maintain commands + get_automation_state/propose_action MCP tools, this completes the CLI + MCP + HTTP control surface. The dashboard Automation tab is owner-led visual work, intentionally separate.

Tests

Repo-scoping, ordering, since/limit filters + their 400/schema rejections, detail sanitisation (+ null passthrough), the maintainer-access gate, non-numeric targetKey → null pullNumber. Full npm run test:coverage green (branches 97.03%); new lines fully covered.

🤖 Generated with Claude Code

[ghost]

reviewbot — ⏳ Waiting for CI to finish before the gate review.

[superagent-security]

⚠️ Contributor trust inconclusive. Click here for more info: Superagent Dashboard

[superagent-security]

Superagent didn't find any vulnerabilities or security issues in this PR.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.62%. Comparing base (1c8975f) to head (18c628f).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #937   +/-   ##
=======================================
  Coverage   96.61%   96.62%           
=======================================
  Files         110      110           
  Lines       14954    14988   +34     
  Branches     5404     5414   +10     
=======================================
+ Hits        14448    14482   +34     
  Misses        104      104           
  Partials      402      402           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.