pm-requirements-copilot
Pass
Audited by Gen Agent Trust Hub on Jul 24, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted external data such as discovery transcripts, meeting notes, and informal messages (Slack/email). This creates a surface for indirect prompt injection where malicious content could attempt to influence the generated requirements or schedules. This risk is mitigated by instructions to flag ambiguities and the mandatory requirement for human review before any output is finalized.
- [COMMAND_EXECUTION]: The skill executes local Python scripts to perform deterministic calculations (e.g., RICE/WSJF scoring, Monte Carlo simulations). These scripts are contained within the skill's own directory and do not involve the execution of arbitrary user-supplied or remote code.
- [SAFE]: The skill enforces a gated pipeline architecture that prevents the AI from automatically proceeding through multiple phases. Each transition (e.g., from Clarification to Structuring) requires explicit user confirmation.
- [SAFE]: External integrations via the GitLab MCP are strictly human-in-the-loop. No mutation operations, such as creating or updating issues, are allowed without direct human confirmation for every action.
Audit Metadata