irudd-ts-evaluate
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes external repositories and executes their test suites, which could contain malicious instructions intended to influence the agent's behavior.\n
- Ingestion points: Target repository source code and test files (SKILL.md).\n
- Boundary markers: The agent is instructed to use a separate checkout, an isolated desktop for GUI tests, and must request explicit user permission before creating a draft PR (SKILL.md).\n
- Capability inventory: Includes file system access, shell command execution for the Archguard CLI and build tools, and execution of the target project's Vitest suites (references/mutation-campaign.md).\n
- Sanitization: Analysis results are submitted via a draft PR for human review; instructions emphasize mapping requirements to supported rules rather than arbitrary execution (references/capabilities.md).\n- [COMMAND_EXECUTION]: The skill utilizes several CLI tools including Archguard, git, and the GitHub CLI (gh) to perform project analysis and report results (SKILL.md, references/capabilities.md).\n- [DYNAMIC_EXECUTION]: The skill involves building the Archguard CLI from source code and executing experimental analysis scripts (runner.py, analyze.py) provided in the tool's repository (references/capabilities.md, references/mutation-campaign.md).\n- [EXTERNAL_DOWNLOADS]: The skill downloads and builds the Archguard tool from the vendor's GitHub repository (alundgren/irudd-ts).
Audit Metadata