The Agent Skills Directory

[SAFE]: The skill functions as a structured guide for manual and automated security reviews. It does not initiate network connections, download external scripts, or access sensitive system files for exfiltration.
[SAFE]: The repository URL (https://github.com/312362115/claude) aligns with the vendor context provided for the author '312362115', representing legitimate project documentation.
[SAFE]: Although the skill is designed to ingest and process untrusted external code for auditing, which is a surface for indirect prompt injection, this behavior is central to its stated purpose as a security tool. The instructions mitigate risks by encouraging an adversarial mindset ('think like an attacker') during the evaluation process.
Ingestion points: Project directories and files (e.g., package.json, requirements.txt, source code) identified during the scope recognition phase in SKILL.md.
Boundary markers: No specific delimiters or 'ignore' instructions are defined for the ingested code content.
Capability inventory: The skill utilizes file reading for analysis and file writing for generating audit reports in the docs/audits/ directory (SKILL.md).
Sanitization: No explicit sanitization or escaping of audited content is mentioned before it enters the agent's context.

security-audit