mck-ppt-design
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes the 'openclaw' CLI tool via subprocess in its example scripts to deliver generated presentations. This operation is limited to handling the path of the generated PPTX file.
- [EXTERNAL_DOWNLOADS]: Fetches AI-generated cover images from the Tencent Hunyuan API and relies on the 'rembg' library which may download pre-trained models from public sources. These interactions are confined to official cloud endpoints and standard libraries.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data to populate slide content, presenting a potential injection surface. Evidence chain: (1) Ingestion points: User text enters the pipeline via Stage 1 (brief.md) and Stage 3 (content.json). (2) Boundary markers: Isolated workflow stages and JSON/Markdown file structures define clear transitions. (3) Capability inventory: Includes file system writes for presentation generation and network/CLI access for assets and delivery. (4) Sanitization: Post-rendering checks via an automated QA engine (qa.py) and XML cleanup logic (core.py) validate output integrity.
Audit Metadata