test-validity-review
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes and analyzes external, potentially untrusted code and configuration files, which serves as a surface for instructions embedded in the project data. * Ingestion points: The agent is instructed to read
package.json, project configurations, setup files, and source code (SKILL.md,scripts/discover-test-surface.mjs). * Boundary markers: The workflow defines a specific verdict scale and finding format to structure the agent's output, helping to maintain task focus. * Capability inventory: The agent is directed to identify and run repository-defined scripts and tests using system tools. * Sanitization: There is no explicit sanitization of the audited code; the skill relies on the agent's core safety guardrails and the provided heuristics to handle external data safely. - [COMMAND_EXECUTION]: The skill's workflow encourages the execution of shell commands and scripts found within the target repository being audited. * Evidence: The
SKILL.mdfile explicitly instructs the agent to 'Run available checks when permitted', 'Use repo commands when available', and perform 'targeted tests, full suite, coverage' or 'mutation testing' using commands discovered in the project's metadata.
Audit Metadata