code-workflow
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project instructions, neighbor files, and code diffs which could contain malicious instructions designed to influence the agent. 1. Ingestion points: SKILL.md (Reading project instructions, neighbor files, and code diffs). 2. Boundary markers: No specific delimiters or instructions to ignore embedded prompts in processed data are defined. 3. Capability inventory: SKILL.md (Executing test suites and running code changes). 4. Sanitization: No sanitization or validation of the ingested content is described.
- [COMMAND_EXECUTION]: The skill instructions mandate running and rerunning tests during development. This involves executing code from the local project, which is a standard behavior for development skills but carries inherent risk if the project files contain malicious code.
Audit Metadata