verification-and-quality-assurance
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze untrusted external content, including codebase files, task outputs, and agent logs, to generate truth scores and quality metrics. * Ingestion points: The skill accesses files via Read, Glob, and Grep tools, including searching the broad path D:\Projects*. * Boundary markers: The instructions do not specify the use of delimiters or 'ignore' commands when the agent processes potentially adversarial content from these files. * Capability inventory: The skill utilizes TodoWrite for file modifications and the claude-flow tool for automated 'auto-fix' and 'rollback' operations on the local filesystem. * Sanitization: There is no documented mechanism for sanitizing or validating external inputs before they are processed by the agent.
- [COMMAND_EXECUTION]: The skill instructions frequently invoke the npx claude-flow@alpha command to execute truth scoring, verification, and dashboarding functions.
- [EXTERNAL_DOWNLOADS]: The reliance on npx leads to the download and execution of the claude-flow package from a public registry at runtime.
- [DYNAMIC_EXECUTION]: The 'auto-fix' functionality allows the agent to dynamically generate and apply code modifications to the codebase, which could be exploited if malicious patterns are injected into the input data.
Audit Metadata