Skill: Security Review

Purpose

Find security risks before a change is accepted.

Inputs

• Change summary or diff.
• Feature docs.
• Architecture and security docs.
• Test results.

Required Reading

• docs/20-security/SECURITY_MODEL.md
• docs/20-security/THREAT_MODEL.md
• docs/10-architecture/FILE_WRITE_POLICY.md
• docs/60-engineering/ENGINEERING_STANDARDS.md
• docs/ai/MCP_STRATEGY.md

Output Files

• Relevant feature REVIEW.md
• Relevant feature COMPLETION_REPORT.md
• Security docs, if the accepted behavior changes.

Process

1. Identify changed trust boundaries.
2. Check path validation, overwrite policy, symlink policy, and dry-run behavior.
3. Check dependency, package, template, and preset risk.
4. Check for network, telemetry, secrets, .env, cloud, AI API, or runtime MCP behavior.
5. Check tests for security-sensitive behavior.
6. Classify findings as blockers, risks, or documented acceptable tradeoffs.

Stop Conditions

Stop and request human decision if:

• Runtime network, telemetry, cloud, MCP, AI API, auth, secrets, storage, or file write behavior

changes without ADR or security review.

• Existing files can be overwritten by default.
• Writes can escape the project root.
• Secrets could be read, logged, or generated into docs.
• Engineering standards are bypassed for secrets, dependencies, migrations, tests, or completion

evidence.

Quality Bar

• Findings are specific and actionable.
• Security claims cite docs or test evidence.
• Remaining risks are explicit.