blast-radius
Pass
Audited by Gen Agent Trust Hub on Aug 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill explicitly instructs the agent to perform dynamic code execution to verify safety claims. Step 4 directs the agent to 'run it' (a script or test), and Step 5 mandates 'Write a script or test that runs the real code, run it, and paste what happened.' This establishes a pattern of runtime script generation and execution.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it processes untrusted external data and uses it to drive high-capability actions.
- Ingestion points: The agent is directed to read code diffs, PR contents, symbols, and external library source code (Step 1, Step 3).
- Capability inventory: The agent has the capability to write files, execute shell commands, and run Python/Node.js scripts to perform tests (Step 4, Step 5).
- Boundary markers: The instructions lack explicit boundary markers or warnings to the agent to ignore instructions embedded within the code being analyzed.
- Sanitization: While the skill advises stripping 'anything private before it goes anywhere public,' there are no instructions for sanitizing or escaping the content of the diffs before interpolating them into generated test scripts.
Audit Metadata