The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a shell script evals/run-static-checks.sh intended for developer-side static verification. This script executes a local Python validator (check-skill-static.py) to ensure the skill's structure and registries remain intact.
[DATA_EXFILTRATION]: The skill is designed to read local codebase files for auditing purposes and write findings to local directories such as docs/audits/ or audit-artifacts/. No unauthorized network communication or external data transmission was observed.
[PROMPT_INJECTION]: The instructions provide clear boundaries and negative test cases to prevent the agent from performing tasks outside its scope (e.g., designing new systems or refactoring code). No bypass or override patterns are present.
[SAFE]: All external references in the documentation and skill.json point to established architectural patterns and well-known academic or industry sources (e.g., Robert C. Martin, Martin Fowler, Alistair Cockburn).

architecture-audit