blast-radius

Pass

Audited by Gen Agent Trust Hub on Aug 19, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill's core workflow involves the agent ingesting and analyzing external, untrusted data in the form of code diffs and pull requests.
  • Ingestion points: Step 1 explicitly directs the agent to read the change diff and use external tools to pull PRs and commits.
  • Boundary markers: The instructions lack boundary markers or specific guidance to ignore potential instructions embedded within the code comments or documentation being analyzed.
  • Capability inventory: The skill requires the agent to have script execution capabilities to fulfill Step 5, which involves running code to prove safety facts.
  • Sanitization: There is no evidence of sanitization or validation of the untrusted code before the agent processes it and uses it to generate verification scripts.
  • [COMMAND_EXECUTION]: The skill encourages the dynamic creation and execution of code within the local environment.
  • Evidence: Step 5 ("Write a script or test that runs the real code, run it") prompts the agent to generate and run arbitrary code based on its analysis of the current project. This presents a risk if the code being analyzed contains malicious logic designed to exploit the agent's execution environment during the verification phase.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 19, 2026, 05:43 PM
Security Audit — agent-trust-hub — blast-radius