security-incident-response

Pass

Audited by Gen Agent Trust Hub on May 13, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads a specialized security scanner from the official Snyk Labs GitHub repository to perform deep analysis of GitHub Actions workflows. Snyk is a well-known security research organization, and this behavior is consistent with the skill's primary purpose.
  • [PROMPT_INJECTION]: As a security analysis tool, the skill is designed to ingest and process untrusted external content, including source code, configuration files, and vulnerability advisories, which constitutes an indirect prompt injection surface.
  • Ingestion points: The agent reads project manifest files (e.g., package.json, pyproject.toml, go.mod), CI/CD workflow definitions (.github/workflows/*.yml), and external security research text into its context.
  • Boundary markers: The instructions guide the agent to distinguish confirmed exposure from unverified evidence, although it does not specify technical delimiters (like XML tags) for the ingested data.
  • Capability inventory: The skill possesses capabilities for file system discovery (find), file reading (cat, rg), network access via the GitHub CLI (gh api), and the ability to execute various local security auditing binaries.
  • Sanitization: The skill explicitly mandates safety-first approaches, such as using --ignore-scripts during dependency audits and prioritizing static inspection over the execution of untrusted project code.
Audit Metadata
Risk Level
SAFE
Analyzed
May 13, 2026, 06:37 PM