skill-install
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from GitHub repositories which poses a risk of indirect prompt injection designed to bypass safety checks or manipulate the installation process.\n
- Ingestion points: GitHub repository content and file data (SKILL.md, scripts) processed in Step 4.\n
- Boundary markers: The skill utilizes a dedicated security scan prompt defined in
references/security_scan_prompt.mdto guide the analysis.\n - Capability inventory: The skill uses the Write tool to create files and directories in the user's
~/.claude/skills/directory (Step 7).\n - Sanitization: Employs an LLM-based security scan and user confirmation for warnings, though these measures do not fully eliminate the risk of sophisticated injection attacks embedded in external content.\n- [EXTERNAL_DOWNLOADS]: Fetches skill configuration and scripts from GitHub's official API and raw content domains.\n
- Details: The workflow downloads content from
api.github.comandraw.githubusercontent.comas part of the installation process.
Audit Metadata