diagnose
Pass
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
Bashtool to perform local file searches (grep) and retrieve issue data from GitHub (gh issue view). These actions are consistent with the stated purpose of bug diagnosis. - [DYNAMIC_EXECUTION]: The skill dynamically generates prompts for subagents using the
Tasktool. It explicitly restricts these subagents to a read-only investigation mode to prevent unintended modifications to the codebase. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources (user input and GitHub issues) by writing it to local session files. 1. Ingestion points: Symptoms are pulled from user input or
gh issue view. 2. Boundary markers: Data is structured within Markdown headers in the session file. 3. Capability inventory: Subagents have access to file system and task management tools. 4. Sanitization: No specific content sanitization is performed on the ingested text. - [SAFE]: The skill maintains its persistent state in a dedicated directory that is explicitly gitignored, following best practices for local state management.
Audit Metadata