spec-loop-plan-task
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: No override markers, bypass instructions, or attempts to extract system prompts were detected. The instructions follow a professional and procedural tone focused on project management.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill does not access sensitive files (e.g., .ssh, .aws) and does not perform any network operations. It manages internal project task files in specific directories (backlog, in-progress, review, done).
- [REMOTE_CODE_EXECUTION]: No remote scripts are downloaded or executed. The skill references other local documentation and skills using relative paths, which is standard for modular agent skills.
- [DYNAMIC_CONTEXT_INJECTION]: No load-time shell execution patterns (
!command) were found in the core SKILL.md file. Commands likerg(ripgrep) are mentioned as tools for the agent to search task files during active sessions, which is benign in this context.
Audit Metadata