stitch-loop
Pass
Audited by Gen Agent Trust Hub on Apr 14, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill implements an autonomous development loop where the agent reads the next task from a local file.
- Ingestion points: The task instructions and visual styles are read from
.stitch/next-prompt.mdand.stitch/DESIGN.md. - Boundary markers: No explicit delimiters or safety instructions are defined to separate the task content from the agent's logic.
- Capability inventory: The agent can use
Bash,Write, and specialized MCP tools for code generation and browser automation. - Sanitization: No validation or filtering is applied to the baton file content before it is passed to the generation tool.
- [COMMAND_EXECUTION]: Instructs the agent to use the
Bashtool to execute local commands such asnpx serveto host a preview of the generated site. - [EXTERNAL_DOWNLOADS]: Fetches HTML code and screenshots from the download URLs provided by the Stitch MCP server to save them into the local repository.
Audit Metadata