The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The documentation and examples suggest downloading an OpenAPI specification from a local development server using curl. While flagged by automated scanners, this command (curl ... -o openapi.json) only saves the file to disk and does not pipe it to an interpreter for execution. The target 127.0.0.1 is the local machine, which is a standard practice for development workflows.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from local files like openapi.json, firestore.rules, and firestore.schema.json. These files could potentially contain malicious natural language instructions designed to influence the agent's behavior when it reads the parsed output.
Ingestion points: openapi.json, firestore.rules, firestore.schema.json (read via scripts/sync_openapi.py and scripts/fetch_firestore_schema.py).
Boundary markers: The scripts do not use explicit delimiters or "ignore instructions" warnings when outputting data to the agent's context.
Capability inventory: The skill has permissions to read and write files within the project root. It does not perform network operations or execute arbitrary subshells.
Sanitization: sync_openapi.py uses regex to sanitize class names, but does not sanitize description fields or other metadata that might be included in the output.

dart-type-bridger