maintain-verification-skill
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of project source code during the 'Source wave' phase to evaluate feature implementation.
- Ingestion points: The skill reads project-level source files and feature map documentation from the local filesystem.
- Boundary markers: The instructions attempt to mitigate risk by specifying that subagents responsible for reading source code are read-only and prohibited from driving the application or editing files.
- Capability inventory: The main agent possesses the ability to drive live application sessions (servers, UIs, CLIs), modify files within the verification skill's directory, and initiate Pull Requests.
- Sanitization: No explicit sanitization, validation, or escaping of the content read from the source code is specified.
- [COMMAND_EXECUTION]: The skill explicitly utilizes 'harness scripts' to perform live verification of features. While these scripts are intended to be part of the local verification infrastructure, their execution involves running code on the host machine to drive and 'doctor' the application state.
Audit Metadata