request-blocks-review

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill manages a standard GitHub PR workflow involving an external integration ('Blocks').
  • [EXTERNAL_DOWNLOADS]: The instructions explicitly mention installing a dependency: npx skills add crissmoldovan/agent-skills --skill blocks. This targets the author's own repository (crissmoldovan) and includes a mandatory instruction to 'Never install it silently,' requiring user approval.
  • [COMMAND_EXECUTION]: The skill uses bash to interact with GitHub via the gh CLI and the blocks skill. These operations are scoped to pull request management and implementation verification (running local tests).
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data (GitHub PR comments and review findings). However, it includes strict boundary logic, requiring the agent to 'Record the head SHA' and 'Compare against the latest request baseline' to prevent accepting outdated or malicious review data. Findings are categorized by severity, and the agent is instructed to ask the user before making judgment calls.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 10:08 AM
Security Audit — agent-trust-hub — request-blocks-review