Security Code Review Agent Skill

---
name: defensive-code-review
description: Review code changes for plausible security regressions. Use for authentication, authorization, payments, uploads, webhooks, secrets, dependencies, or other trust-boundary changes.
---

# Defensive Code Review

## Review method

1. Define the changed assets, actors, entry points, trust boundaries, and attacker-controlled inputs.
2. Trace data from input through validation, authorization, storage, output, and logs.
3. Check server-side authorization for every sensitive read and write, including tenant ownership.
4. Review injection risks across SQL, shell, templates, URLs, headers, file paths, and model/tool prompts.
5. Check authentication lifecycle: creation, rotation, expiry, revocation, replay, and recovery.
6. Inspect secret handling, error messages, telemetry, and client bundles for accidental disclosure.
7. Review external calls for SSRF, redirect abuse, signature verification, timeouts, retries, and idempotency.
8. Review uploads for type confusion, size limits, storage isolation, path traversal, and unsafe processing.
9. Review new dependencies and CI changes for supply-chain and permission expansion.

## Findings

- Report only findings supported by a concrete code path.
- For each finding include severity, affected file/line, exploit scenario, impact, and minimal remediation.
- Separate confirmed vulnerabilities from hardening suggestions.
- If no issue is found, state the areas reviewed and remaining uncertainty.

## Safety

- Keep testing defensive and scoped to the authorized repository.
- Do not access production data, exploit live systems, or reveal secrets.