prompting
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- SAFE (SAFE): The skill consists entirely of Markdown documentation and workflows. There are no executable scripts (Python, Bash, or Node.js) included in the provided files.
- DATA_EXPOSURE (SAFE): The skill uses internal path variables like
${PAI_DIR}for documentation cross-referencing, which is standard for relative pathing in these environments. No hardcoded secrets, credentials, or sensitive system paths were detected. - EXTERNAL_DOWNLOADS (SAFE): Includes a link to a legitimate Anthropic engineering blog post for reference. No automated downloads, package installations, or remote script executions are performed.
- PROMPT_INJECTION (SAFE): The instructions focus on defensive prompt engineering (clarity, structure, and constraints). There are no malicious override patterns, role-play injections, or instructions to bypass safety filters.
Audit Metadata