qa
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill constructs shell commands using variables derived from external tools (git, gh) and user arguments. While these are typical for development tools, maliciously crafted filenames or branch names could potentially lead to command injection in loops that process file lists (e.g., Step 4, Diff-Aware Mode).
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes local source code and test files to determine dependencies and coverage. This processing of untrusted local data provides a surface for indirect prompt injection, where malicious instructions hidden in code comments or strings might attempt to influence the agent's behavior during the QA process.
- Ingestion points: Local source code and test files read via grep and file system searches.
- Boundary markers: None explicitly mentioned in the processing logic.
- Capability inventory: Execution of shell commands (git, find, grep, pnpm, npm, cargo, go) and automated Git commits for fixes.
- Sanitization: No explicit sanitization or escaping of file content or metadata is defined before it is used in logic or reports.
Audit Metadata