The Agent Skills Directory

[SAFE]: The skill performs legitimate synthesis of conversation and codebase data into documentation. No malicious patterns such as remote code execution, persistence, or data exfiltration were detected.
[INDIRECT_PROMPT_INJECTION]: The skill demonstrates an attack surface for indirect prompt injection as it ingests untrusted data from the codebase and conversation history. Ingestion points: codebase files and conversation context (SKILL.md). Boundary markers: None present; the instructions do not specify delimiters to separate data from instructions. Capability inventory: Repo exploration and publishing to an issue tracker (SKILL.md). Sanitization: None present in the instructions.

to-spec