[ScriptedAlchemy]

Merge #4478 first, followed by #4526, #4510, #4524, #4489, #4485, #4527, #4528, #4530, #4531, #4529, #4534, and #4541.

This PR is number 7 of 13, based on #4485. Merge #4528 next.

Use merge commits. After merging a parent PR, change its child's base to master before merging the child. Auto-merge is disabled.

Flat YAML sometimes uses the HTTP-header parser. That parser checked field names for known sensitive meanings and inspected values, but it did not detect credentials in the keys themselves. A credential-shaped key could leave its value exposed. A leading UTF-8 BOM also prevented JSON and TOML from reaching structural checks.

The line-format sanitizer now detects credentials in keys, including keys with empty values. It uses the existing detectors and typed quarantine result.

JSON and TOML processing removes one leading BOM before parsing, finding spans, rewriting, and checking the result. Text output restores the BOM. Admission limits and receipt identity still use the original bytes. Shared JSON parsing and LCM JSON detection accept BOM-prefixed input while preserving duplicate-key refusal and LCM's larger JSON allowance.

The code-source implementation revision and LCM detector revision advance so retained generations and completed rescans cannot continue to certify content accepted by the old checks. The in-place JSON and TOML key redaction and reparse checks from #4515 remain.

Fixes #4519.

Validation

Canonical Bazel validation on the integrated parent tree is recorded below.

• Integrated parent head: 2af7f6d6f4a0f2badc6951a52749bd22f97ef1ef.
• Privacy unit tests: 161 passed; search-cache retention: 1 passed; affected Rust formatting passed. Fresh Bazel invocation 29b6c73b-ec84-4468-8a95-3fdc458af26b.
• Scoped Clippy for both targets passed: ada2671a-ed9b-42ef-b07b-8c3f00d94464.
• These checks used canonical Bazel targets on Linux. Hosted checks remain independently required.

Structural identity admission correction

Fleet validation on descendant #4528 exposed Kimi and Vibe prefixed-session admission failures. The canonical structural-ID helper now digests quarantined native identifiers using its existing digest format, while sanitizer errors and scan limits still refuse admission. Public and previously protected identities remain stable.

Fix: 699c648217. Controlled Kimi BEFORE failed (0a07923d-9131-47ae-b88a-f63cf501fb8a); the new focused structural-ID regression also failed before the helper change (8c478185-6935-4c54-8f61-515826619192). AFTER: all 163 privacy tests and 633 session tests passed, including both original failing host journeys, plus formatting (84e89e56-2aaa-403c-a04c-3a27b2f41436). Scoped Clippy passed (8e707247-fbe2-415c-bda9-91478a20148f). No host assertions or admission limits changed.

[changeset-bot]

⚠️ No Changeset found

Latest commit: 6b5a992

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.