The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill provides instructions and examples for using a Python script (upload_to_stitch.py) that accepts an --api-key parameter. Guiding an AI agent to pass sensitive credentials as command-line arguments is insecure, as these values can be exposed in process listings, logs, or command history.
[COMMAND_EXECUTION]: The skill relies on the execution of a local Python script located at stitch-skills/plugins/stitch-design/skills/upload-to-stitch/scripts/upload_to_stitch.py to handle design system uploads. While this appears to be a vendor-provided tool, it represents a core capability of the skill to execute arbitrary code on the host system.
[DATA_EXFILTRATION]: The skill uses read_url_content and web_fetch to retrieve project assets (HTML/Metadata) from remote URLs and subsequently uploads processed data to the Stitch platform. While this is part of the design management workflow, it establishes a data flow from external sources to the Stitch API.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8). It retrieves HTML code and project metadata from external URLs (e.g., screenshot.downloadUrl, htmlCode.downloadUrl) using read_url_content.
Ingestion points: External HTML content is fetched via read_url_content based on project metadata.
Boundary markers: The instructions do not specify any delimiters or warnings to ignore instructions embedded within the fetched HTML code.
Capability inventory: The skill has access to Bash, Write, and Stitch-specific tools, allowing it to modify files or execute commands based on instructions it might process.
Sanitization: There is no mention of sanitizing or validating the fetched HTML content before it is processed by the agent.

stitch::manage-design-system