start
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes 'which' commands via the Bash tool to verify the presence of security scanners like semgrep, gitleaks, and trivy on the host system (SKILL.md Step 4).\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze project files, which constitutes an attack surface for indirect prompt injection if those files contain malicious instructions.\n
- Ingestion points: Project manifest files (e.g., package.json, requirements.txt, Dockerfile) and source code scanned via Grep (SKILL.md Step 1, 2, 3, 5).\n
- Boundary markers: None; the skill reads and processes file content directly for keyword detection.\n
- Capability inventory: Limited to Glob, Grep, Read, and Bash (for binary detection). No file-write beyond a local cache or network access is requested.\n
- Sanitization: None; the skill relies on the agent to interpret findings from raw file content.\n- [SAFE]: The skill operates as a diagnostic utility within the local project environment, focusing on metadata and configuration discovery to facilitate security best practices.
Audit Metadata