checkpoint

Pass

Audited by Gen Agent Trust Hub on Feb 23, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses shell commands to manage Git operations and file system locking.
  • Evidence: mkdir for lock creation, git status, git add, git commit, and git push are explicitly mentioned or implied.
  • [PROMPT_INJECTION]: The 'Validate' step (Step 3) introduces an indirect prompt injection surface by instructing the agent to extract and potentially execute commands found in untrusted repository files.
  • Ingestion points: Repository documentation and configuration files including README.md, CONTRIBUTING.md, build scripts, and CI notes.
  • Boundary markers: None defined to isolate the discovered instructions from the agent's execution context.
  • Capability inventory: The skill implies the ability to execute "validation instructions" found in the repository.
  • Sanitization: No sanitization or verification of the discovered commands is mentioned.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 23, 2026, 09:54 AM