The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources that could contain malicious instructions designed to influence the agent's behavior during the code generation process.
Ingestion points: Design JSON retrieved via stitch:get_screen and the source HTML downloaded via scripts/fetch-stitch.sh (as seen in SKILL.md).
Boundary markers: None identified; the agent is instructed to directly process the external HTML to extract styles and configuration.
Capability inventory: The agent has access to Bash, Write, Read, and web_fetch, allowing for file system modifications and network access.
Sanitization: There is no explicit instruction to sanitize or validate the content of the external HTML or metadata before the agent uses it to generate component logic.
[COMMAND_EXECUTION]: The skill relies on executing a local shell script to handle networking tasks, which introduces a dependency on the system environment.
Evidence: SKILL.md instructs the agent to use the Bash tool to run bash scripts/fetch-stitch.sh "[htmlCode.downloadUrl]" "temp/source.html".
[EXTERNAL_DOWNLOADS]: The skill fetches content from dynamic URLs provided by the Stitch MCP server to retrieve design assets.
Evidence: scripts/fetch-stitch.sh contains a curl command used to download HTML and screenshot assets from Google Cloud Storage or other remote domains provided in the tool metadata.

react-components