pmg-engine
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input in the form of presentation specifications (JSON) to generate PowerPoint files.
- Ingestion points:
SKILL.mdspecifies building presentations frompresentation_spec.jsonvia thepmg buildcommand and Python API. - Boundary markers: Absent; the documentation does not describe using delimiters or warnings to ignore instructions inside the input data.
- Capability inventory: The skill can write files (.pptx, .mp4) and initiate network connections to remote worker URLs for distributed QA.
- Sanitization: Absent; no mention of content filtering or schema enforcement for the input JSON specification.
- [COMMAND_EXECUTION]: The skill utilizes a CLI tool (
pmg) to perform environment diagnostics, file generation, and remote validation tasks. - Evidence: CLI examples provided in
SKILL.mdsuch aspmg doctor,pmg build, andpmg qa.
Audit Metadata