feat(engine): close blacklisted contributors' PRs ahead of merit (#1425)

[JSONbored]

Summary

The engine disposition for the anti-abuse contributor blacklist (#1425). A PR from a blacklisted
login short-circuits the agent maintenance planner to a deterministic label + close, ahead of all
merit / CI / gate / AI analysis — with a sanitized close comment and no AI call.

Stacked on #1430 (the blacklistLabel config layer) — review/merge that first.

Behavior

• Short-circuits and wins over the normal gate. Before any CI/merit/AI step, a blacklist match
  produces [label, close] and returns — the PR is never merit-reviewed or auto-merged.
• Configurable label. Uses the resolved blacklistLabel (default slop); works regardless of the
  label a repo sets — nothing hard-coded.
• No AI call. The AI review is skipped for a blacklisted author (the close is deterministic).
• Public-safe. The maintainer's reason is run through the existing public-safe sanitizer, so a
  private term can never leak into the PR thread.
• Respects existing controls. Honors the autonomy dial and agentPaused / agentDryRun exactly
  like every other agent action; closeKind: "blacklist" is exempt from AI refutation (like the
  linked-issue hard rule). The owner and automation bots are never auto-closed.

Scope

• Narrow, in wantedPaths; no site/ / CNAME / lovable; no changelog edit.
• Per-repo list resolved now; the shared/global union primitive (mergeContributorBlacklists) is in place for when the global table lands.
• No secrets / wallets / hotkeys / trust scores / private values; logins are public data only.

Validation

• npm run typecheck
• npm run test:coverage — full suite green; 100% of changed lines and branches covered.
• Planner unit tests (short-circuit precedence, owner/bot exemption, autonomy, label config, sanitization) + an end-to-end queue test proving labeled + closed + no AI call + no merge for a banned author.

Safety

• Close comment goes through sanitizePublicComment; a forbidden term in the reason is asserted absent.
• No new auth/CORS surface; deterministic path, no model call.

Advances #1425.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.40%. Comparing base (eb90475) to head (83099b9).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@                   Coverage Diff                   @@
##           claude/blacklist-label    #1431   +/-   ##
=======================================================
  Coverage                   95.39%   95.40%           
=======================================================
  Files                         193      193           
  Lines                       20932    20945   +13     
  Branches                     7566     7573    +7     
=======================================================
+ Hits                        19969    19982   +13     
  Misses                        383      383           
  Partials                      580      580           

Files with missing lines	Coverage Δ
src/queue/processors.ts	88.31% <100.00%> (+0.02%)	⬆️
src/settings/agent-actions.ts	92.79% <100.00%> (+0.71%)	⬆️
src/settings/contributor-blacklist.ts	100.00% <100.00%> (ø)

🚀 New features to boost your workflow:

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