Summary

• scanPatch (the secret-scan analyzer's only live entry point, via scanSecrets/secret/descriptor.ts) matched each added diff line independently, so a real credential split across two added lines via string concatenation (const a = "AKIA..."; const b = a + "REST";) never appeared as a single matchable token on either line and evaded every RULES regex.
• Extracts quoted string-literal contents per added line and, when a line doesn't already match on its own, joins the immediately-preceding line's last literal with the current line's first literal and tests that too — downgraded to medium confidence since a joined pair is a heuristic, not a direct contiguous match.
• Bounded to consecutive added lines only: a hunk boundary, a context line, or a removed line resets the join window, so this only ever catches the realistic "two sequential variable assignments" shape, not an unbounded cross-file join.
• Existing per-line detection is untouched and unweakened.

Scope

• The PR title follows type(scope): short summary Conventional Commit format, for example fix(api): restore profile access checks.
• This PR is focused and does not mix unrelated backend, UI, MCP, docs, dependency, and deploy changes.
• This follows CONTRIBUTING.md and does not reintroduce GitHub Pages, VitePress, site/, or CNAME.
• I linked an issue, or this is small enough that the summary explains why an issue is not needed.

Validation

• git diff --check
• npm run actionlint
• npm run typecheck (root — unaffected, no src/** changed)
• npm run test:coverage — N/A for this PR's own diff: review-enrichment/** is a separate workspace excluded from the Codecov src/** patch gate (own npm run rees:test runner instead, see below)
• npm run test:workers
• npm run build:mcp
• npm run test:mcp-pack
• npm run ui:openapi:check
• npm run ui:lint
• npm run ui:typecheck
• npm run ui:build
• npm audit --audit-level=moderate — 0 vulnerabilities
• New or changed behavior has unit/integration tests for new branches, fallback paths, and sanitizer boundaries

Also ran npm run rees:test (the analyzer package's own gate, part of test:ci): 374/374 passing, including a new review-enrichment/test/secret-scan.test.ts (9 tests) covering the pre-existing single-line rules plus 5 new tests for the cross-line join: the concatenation-split regression case itself, a context-line break, a hunk-boundary break, no-double-report when a line already matched directly, and a false-positive guard for two unrelated short literals. npm run db:migrations:check also unaffected/green.

Safety

• No secrets, wallet details, hotkeys, coldkeys, user PATs, private keys, raw trust scores, private rankings, or private maintainer evidence are exposed. Test fixtures build fake secret-shaped strings from split fragments at runtime rather than as contiguous literals, specifically so the fixture file itself doesn't trip GitHub push protection the way a real leaked credential would.
• Public GitHub text stays sanitized, low-noise, and does not imply compensation guarantees or optimization tactics.
• Auth, cookie, CORS, GitHub App, Cloudflare, or session changes — N/A, this is an enrichment analyzer, not an auth path.
• API/OpenAPI/MCP behavior is updated and tested where needed. No OpenAPI/MCP surface changed.
• UI changes — N/A, no UI changed.
• Visible UI changes — N/A, no UI changed.
• Public docs/changelogs are updated where needed; changelogs are only edited for release-prep PRs. No changelog edit.

Notes

• Found via an adversarial audit of the review engine's security-relevant analyzers (which gate what counts as risky code in a PR). scanAddedLinesForSecrets (a second, unused export in the same file with the same per-line-only gap) was left untouched — it has zero callers anywhere in the codebase, so fixing it would add untestable dead-code coverage; out of scope for this fix.