stitch-loop
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill operates by processing content from external 'baton' files which could potentially contain untrusted instructions.
- Ingestion points: The skill reads
.stitch/next-prompt.md,.stitch/SITE.md, and.stitch/DESIGN.mdto determine next steps for the agent. - Boundary markers: The skill uses standard YAML frontmatter for metadata but lacks specific security delimiters to isolate instruction content from the rest of the file.
- Capability inventory: The skill utilizes Stitch-specific MCP tools and file system operations to move and verify generated HTML assets.
- Sanitization: No explicit sanitization or filtering of the text content within the baton files is performed.
Audit Metadata