The Agent Skills Directory

[SAFE]: The skill is purely analytical and instructional, focusing on generating plans for mechanical code enforcement rather than executing commands or modifying the system directly.\n- [SAFE]: It implements a 'Judgment Verification Pass' to ensure that all proposed rules are based on stable mechanical signals, effectively filtering out subjective AI hallucinations or taste-based reviews.\n- [SAFE]: Strict operational constraints are established in the contract, explicitly forbidding the installation of dependencies or modification of CI configurations without direct user authorization.\n- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it processes untrusted data from Sentry, GitHub, and session transcripts. However, the risk is mitigated by the skill's lack of exploitable capabilities (no shell access, network operations, or file-write permissions). Ingestion points: evidence sources defined in SKILL.md. Boundary markers: Absent. Capability inventory: No dangerous tools or subprocess calls enabled. Sanitization: Absent.

odie