The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script scripts/upload_to_stitch.py to handle Base64 encoding and transmission of files, which avoids the token limit constraints of the AI agent's primary output channel.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to https://stitch.googleapis.com to upload assets. This is a well-known service endpoint belonging to a trusted organization.
[DYNAMIC_EXECUTION]: The skill's documentation and automated tests use importlib.util to dynamically load and validate the local upload_to_stitch.py script. This execution is limited to a static, local file path.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided HTML and Markdown files, which could serve as a vector for indirect prompt injection if the files contain instructions meant to influence the agent or downstream systems. 1. Ingestion points: File content is read via the --file-path parameter in scripts/upload_to_stitch.py. 2. Boundary markers: The skill relies on manual user desensitization as instructed in the SKILL.md and does not implement programmatic content delimiters. 3. Capability inventory: The skill has network write access to the Stitch API for asset creation. 4. Sanitization: No automated content filtering or sanitization is applied to file data before it is encoded and transmitted.

stitch-upload-to-stitch