pm-state
Pass
Audited by Gen Agent Trust Hub on Jul 2, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses localized file system paths (
docs/pm-context/,.loop/) to read documentation and write diagram outputs. These are standard documentation workflows for AI agents and do not involve system-level command execution or shell injection patterns. - [PROMPT_INJECTION]: The instructions contain strict constraints and self-correction protocols (Thinking Loop Step 6) to ensure the AI remains grounded in the provided
PMContext. There are no attempts to bypass safety filters or ignore prior instructions. - [DATA_EXFILTRATION]: No network tools are requested or used. The skill's data flow is entirely local, moving from a local input file (
pm-context.md) to a local output file (state.md). - [EXTERNAL_DOWNLOADS]: The skill mentions external documentation (Mermaid/Wikipedia) for reference only; it does not attempt to download or execute external scripts or packages.
- [REMOTE_CODE_EXECUTION]: The skill generates Mermaid syntax (text-based diagrams). It specifically instructs the AI to avoid using zero-width characters in Mermaid blocks and uses static logic to handle state transitions, presenting no risk of dynamic code execution.
Audit Metadata