The Agent Skills Directory

[SAFE]: The skill is composed of instructional Markdown files providing guidelines for auditing and writing developer documentation. It does not include any executable scripts, environment configuration files, or tools that could be used for malicious purposes.- [INDIRECT_PROMPT_INJECTION]: The skill processes external documentation provided by the user (Category 8a). However, it implements a rigorous audit protocol and scoring system that mandates factual verification against source code (Category 8c), effectively mitigating risks associated with malicious instructions embedded in the documentation being audited.- [PROMPT_INJECTION]: Deterministic detections for concealment patterns were identified as false positives. The instructions to the agent to 'never include a command you didn't see in code' and to mark unverified facts with 'TODO(verify)' are quality-control measures designed to prevent hallucinations and ensure accuracy, rather than attempts to bypass security filters.

technical-documentation