pre-release-review

Pass

Audited by Gen Agent Trust Hub on Jul 9, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill analyzes untrusted data from git diffs, pull request metadata, and repository-local guidance files (AGENTS.md, CLAUDE.md), which creates a surface for indirect prompt injection.
  • Ingestion points: The workflow involves fetching and reading content from git diff, gh pr view, and project configuration files.
  • Boundary markers: While no technical delimiters are defined for the input data, the skill includes explicit 'non-negotiable rules' instructing the agent to maintain read-only behavior and ignore conflicting instructions in project guidance.
  • Capability inventory: Tool access is limited to git, gh (GitHub CLI), and rg (ripgrep) for inspection purposes.
  • Sanitization: The skill relies on instructional constraints rather than programmatic sanitization of the data being audited.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 9, 2026, 09:08 PM
Security Audit — agent-trust-hub — pre-release-review