feat(sanitizer): configurable secret policy, PEM/JWT detection, honest docs

[chrisleekr]

Why

The secret scanner had two honest problems flagged in review: (1) it only matched ~9 vendor key formats, so a PEM private key or JWT pasted into config flowed straight into the vault, and (2) the docs oversold it as a secret scanner when it is really a known-credential-format guard. There was also no way to exempt a legitimate high-entropy value the base64 redactor would silently corrupt. PR #185 added the [security] config surface (recorded but inert); this PR wires it in and retires the "inert" note.

Third PR in the series (follows #184, #185).

What

• Configurable SecretPolicy resolved from [security] via securityToPolicy, threaded through scanForSecrets (the push-time artifact-body gate) and redactSecretLiterals/sanitizeAndNormalizeJson (per-adapter JSON redaction in claude/codex/cursor/vscode).
  • secretScan: standard (built-in patterns) | strict (+ JWT) | off (disables the artifact-body scan).
  • allowSecretValues: exempts exact literals from detection and base64 redaction (fixes the silent base64 false-positive corruption).
  • redactBase64Values: disable the broad base64 whole-value redaction for configs that legitimately store base64.
• Breadth: PEM private-key headers detected in every mode (unambiguous); JWTs added in strict.
• Fail-safe: off disables the artifact-body scan, but skill-bundle interiors are always scanned at standard (the walker is not policy-threaded), and config set always scans values regardless of off (config is plaintext).
• Honesty docs: architecture/commands/operations now describe the scan as a known-credential-format guard, not a general secret scanner, list exactly what is and isn't covered, and document the tuning knobs.

Security (caught in review)

The first cut of the allowlist used text.match() (first occurrence only): an allow-listed decoy earlier in a file would have masked a real secret of the same shape later — a fail-open leak when allowSecretValues is set. Fixed to iterate every occurrence with matchAll and exempt only exact allow-listed literals; DEFAULT_SECRET_POLICY is frozen against mutation. Default behaviour (no [security]) is byte-identical to before — verified by the unchanged existing sanitizer/adapter suites.

Tests

New "secret policy" unit block + the decoy-masking regression test; end-to-end push tests for secretScan=off (lets a key through), strict (JWT aborts, standard clean), and allowSecretValues (embedded key exempted); adapter test proving the policy is honoured (off → unredacted). Full suite: 936 pass / 0 fail.

CI note

bun test exits non-zero on the per-file coverage floor; CI treats 0-fail as success.

🤖 Generated with Claude Code

[coderabbitai]

Warning

Review limit reached

@chrisleekr, we couldn't start this review because you've reached your PR review rate limit.

More reviews will be available in 5 minutes and 50 seconds. Learn how PR review limits work.

Your organization has used up its prepaid credits, and credit purchases are no longer available. Enable the review add-on in the billing tab to keep reviews running — you're only billed for reviews past your plan's rate limits ($0.25/file).

⌛ How to resolve this issue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based credits.

🚦 How do rate limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan refill rate.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, the refill rate gradually slows as usage increases. The highest same-day bursts are limited more strictly.

Please see our Fair Usage Limits Policy for further information.

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: 5db8f108-3536-4ebe-9675-628fabc6946c

📥 Commits

Reviewing files that changed from the base of the PR and between b9d39ae and 39f96ab.

📒 Files selected for processing (14)

• docs/architecture.md
• docs/commands.md
• docs/operations.md
• src/agents/claude/__tests__/sanitize.test.ts
• src/agents/claude/index.ts
• src/agents/claude/sanitize.ts
• src/agents/codex/index.ts
• src/agents/cursor/index.ts
• src/agents/vscode/index.ts
• src/commands/__tests__/push.test.ts
• src/commands/push.ts
• src/config/schema.ts
• src/core/__tests__/sanitizer.test.ts
• src/core/sanitizer.ts

---

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 and usage tips.}