blast-radius

Pass

Audited by Gen Agent Trust Hub on Nov 16, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content, which could contain malicious instructions. \n
  • Ingestion points: The skill reads diffs, pull request (PR) descriptions, commits, and library source code (SKILL.md Steps 1 and 3). \n
  • Boundary markers: There are no instructions for the agent to use delimiters or ignore instructions within the ingested data. \n
  • Capability inventory: The skill can write and execute scripts, perform shell searches, and access repository symbols (SKILL.md Steps 3 and 5). \n
  • Sanitization: The instructions advise the user to strip private information before public sharing, but no automated sanitization for the model's ingestion is provided. \n- [DYNAMIC_EXECUTION]: The skill explicitly commands the agent to generate and run code to verify safety assumptions. Evidence: SKILL.md Step 5 instructs the agent to "Write a script or test that runs the real code, run it, and paste what happened." \n- [COMMAND_EXECUTION]: The skill utilizes command-line utilities for repository investigation. Evidence: SKILL.md mentions the use of grep and pulling PR data to trace code paths and indirect consumers.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 16, 2026, 07:58 AM