secret-scan
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
trufflehog,gitleaks,gh, andgrepvia subprocess commands. It includes anexecutemode that runs command batches autonomously by default. While restricted to user-provided targets, the instruction to "request approval only for OPSEC-dangerous command batches" gives the agent significant autonomy over standard command execution. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and triage external, untrusted data (GitHub repositories, organization code, and search results).
- Ingestion points: Reads code content and metadata from GitHub (via
gh searchandtrufflehog) and local files (SKILL.md). - Boundary markers: None identified; instructions do not explicitly tell the agent to ignore instructions embedded in the scanned code.
- Capability inventory: Full shell command execution capability (
trufflehog,gitleaks,git clone,grep,mkdir) across SKILL.md. - Sanitization: None; the skill reads raw text matches and code blocks for triage, which could contain malicious prompts targeting the agent's logic.
Audit Metadata