The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes standing source code, which represents a potential ingestion surface for indirect prompt injection. However, the risk is thoroughly mitigated by explicit instructions to the agent to treat all input data as untrusted evidence. Furthermore, the scripts/render-report.sh script implements a comprehensive sanitization pipeline using regular expressions to redact private keys, tokens, emails, and home paths. A residual check ensures that if any sensitive patterns remain, the process fails without writing the report.
[COMMAND_EXECUTION]: The skill utilizes standard system utilities such as git, jq, and find to perform its diagnostic tasks. These commands are executed within the local workspace on explicit targets and are used in a controlled manner that adheres to the principle of least privilege.
[SAFE]: The skill demonstrates a strong commitment to security through its design as a non-authoritative, report-only tool. It explicitly prevents any mutation of source code or external project management artifacts (GitHub, Linear, Plane) without human intervention and separate workflows, thereby maintaining a clear boundary between diagnosis and remediation.

woostack-audit