review-code-security

Scan implemented code for security anti-patterns across the same three axes as /review-plan-security, via Codex through the shared review loop (round-continuing under the 2+1 cap: fix concerns and re-run). Use after implementation to verify the plan's security goals hold in the actual code.

When to Use

Use when the implementation touches external input, secrets handling, or third-party integrations.

Rules

• Report each finding with: file/location, pattern category, and recommended fix.
• Note context for potential false positives (test fixtures, comments, examples).
• For commit-time auto-scanning of hard secrets (AWS/LLM API keys, PEM keys, GitHub/Slack tokens, .env commits), see docs/scan-outbound.md — those are covered automatically.
• Sibling sweep: enumerate functions, logic, or patterns the change touches that belong to a class; flag siblings as MUST / OPTIONAL / NA for same treatment.

Procedure

RCS-1. Resolve <PLANS_DIR> via skills/_shared/resolve-plans-dir.md. Initialize EXTENSIONS_USED=0. RCS-2. Invoke "$AGENTS_MAIN_ROOT/skills/review-code-security/scripts/run-codex-review-loop.sh" (Bash), exporting AGENTS_MAIN_ROOT, SESSION_ID, PLANS_DIR, EXTENSIONS_USED. Pass CTX_SURVEY_CODE, CTX_SURVEY_HISTORY when available. CTX_CONCERNS_LOG is auto-generated by the wrapper from the ledger (concern-ledger render-concerns-log, format review-security-shared) and auto-exported when non-empty — do not set it manually (#2185). Exit-code handling SSOT: skills/_shared/codex-review-loop.md. On exit 3 (codex unavailable): load prior concerns via bash "$AGENTS_MAIN_ROOT/bin/concern-ledger" render-prior --plans-dir <PLANS_DIR> --session-id <session-id> --format review-security-shared; also render the carrier via bash "$AGENTS_MAIN_ROOT/bin/concern-ledger" render-concerns-log --plans-dir <PLANS_DIR> --session-id <session-id> --format review-security-shared and, if the returned path is non-empty and non-zero-size, export it as CTX_CONCERNS_LOG in the security-scanner subagent's environment so rejected concerns are not re-raised. Launch the security-scanner subagent with model: = the model= line of node "$AGENTS_MAIN_ROOT/bin/resolve-role-model" --role reviewer, then re-run the wrapper with --prestaged-report <artifact_path> --prestaged-producer security-scanner --prestaged-exec <COMPLETE|PARTIAL|ABSENT> (mapping scanner status: complete→COMPLETE / partial→PARTIAL / failed→ABSENT); on failed still re-run with --prestaged-exec ABSENT so the round is recorded. The security-scanner JSON prompt sets its dispatch_directive key to the canonical line of the rules/shell-commands.md Subagent Dispatch Prompts section. RCS-3. Triage — per skills/_shared/priority-hierarchy.md, reject a concern only when it directly contradicts a decision already settled in the approved intent.md / outline.md (including their ## Accepted Tradeoffs). Raising a topic the code does not address is never grounds to reject; a reject must cite the specific governing decision. For each rejected concern, record the verdict durably so it does not re-surface next round: bash "$AGENTS_MAIN_ROOT/bin/concern-ledger" reject --plans-dir <PLANS_DIR> --session-id <session-id> --format review-security-shared --id C<N> --reason "<cited decision>" (#2185).