fix(handoff): fail content guard lint on injection warnings

[solomonneas]

Summary

• Make warning-level handoff-injection findings fail brigade handoff lint --content-guard.
• Report egress and injection as separate verdicts in text and JSON output.
• Add lint tests for injection-only failure, egress-only failure, a clean two-axis pass, and non-blocking info findings.

Wiring choice

This uses the in-guard option. --content-guard already calls the canonical scan_handoff_injection_heuristics() detector, so its combined verdict now fails when either egress or injection fails. Info-only discussion remains non-blocking. No second detector was added.

Before and after

Before, an egress-clean injection fixture printed warnings but exited 0:

[ok] content_guard leaks: .../evil-ignore-previous.md clean
  warning: 2 injection heuristic hit(s)
exit: 0

After, each axis has its own verdict and the command exits 1:

[ok] content_guard egress: .../evil-ignore-previous.md clean
[fail] content_guard injection: .../evil-ignore-previous.md 2 injection warnings
exit: 1

The before output is recorded in Brigade receipt 20260726-210033-work-verify-08af8a.

Coordination

Issue #490 owns the model_trials adversarial fixture. This PR does not change model_trials or add another fixture corpus. It uses the existing sample under tests/fixtures/handoff_lint/injection/.

This is a follow-up to #477 and #515, which added the scanner but left warning-level injection findings outside the command verdict.

Closes #555.

Verification

• ./scripts/verify: 4,309 passed, 3 skipped, 82.84% coverage. Brigade receipt 20260726-210442-work-verify-21bcd1.
• pytest tests/test_handoff_injection_lint.py -q: passed. Brigade receipt 20260726-210426-work-verify-98da75.
• Embedded content guard over all six inert attack fixtures: passed. Brigade receipt 20260726-211101-work-verify-f34c42.
• Memory handoff lint with --content-guard: passed. Brigade receipt 20260726-211138-work-verify-5eb8a7.

[coderabbitai]

Important

Review skipped

Auto reviews are limited based on label configuration.

🏷️ Required labels (at least one) (1)

• coderabbit-review

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository: escoffier-labs/coderabbit/.coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 42e969ce-6b6a-434b-a199-f2cd975fbf8b

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/content-guard-injection-crossref

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}