safe-ship
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions involve executing Git commands and project-specific validation tools like linters and test runners, which is standard behavior for developer-focused tasks.
- [INDIRECT_PROMPT_INJECTION]: The agent processes untrusted data from local file contents and command outputs which could theoretically contain instructions designed to influence the agent.
- Ingestion points: Local repository state via
git diff,git log, and validation tool outputs (SKILL.md). - Boundary markers: None explicitly defined in the instructions to separate untrusted output from agent instructions.
- Capability inventory: Execution of arbitrary shell commands for testing, linting, and building; Git operations; file deletion via
trash(SKILL.md). - Sanitization: No explicit sanitization or filtering of external or file content is mentioned.
Audit Metadata