work-on
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The validation script 'scripts/check-single-change-fixture.py' executes 'git' and 'python' commands via 'subprocess.run'. These operations are confined to a temporary directory and are used to verify the skill's procedural instructions against synthetic Git repositories.
- [DYNAMIC_EXECUTION]: The 'scripts/check-single-change-fixture.py' script generates and executes Python code (e.g., 'formatter.py', 'check.py') at runtime to simulate development tasks and verify the accuracy of the skill's logic. These scripts are created from hardcoded templates within the test suite.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as bug reports and analysis documents.
- Ingestion points: 'SKILL.md' (Step 2: Compare supplied investigation/analysis) and 'references/contracts.md' (Handoff envelope).
- Boundary markers: The skill defines a structured 'Handoff envelope' with specific fields for agreement, context, and authority to isolate ingested data.
- Capability inventory: The skill performs file system operations and executes shell commands through its companion scripts ('scripts/check-single-change-fixture.py').
- Sanitization: No explicit string sanitization is described; however, the workflow mandates independent review and verification gates to validate all outputs.
Audit Metadata