The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides instructions for creating web endpoints (e.g., FastAPI, Flask, Django) that ingest untrusted user data via HTTP request bodies and parameters. This establishes an indirect prompt injection surface if the received data is subsequently passed to an LLM without proper sanitization.
Ingestion points: SKILL.md (FastAPI hello and square functions) and references/web-endpoints.md (POST endpoints with Pydantic models).
Boundary markers: Absent from the provided implementation examples.
Capability inventory: The skill demonstrates the use of subprocess.Popen to manage servers and supports full web framework integration.
Sanitization: Input validation or sanitization logic is not explicitly addressed in the code snippets.
[COMMAND_EXECUTION]: The skill demonstrates the use of subprocess.Popen to initiate local HTTP servers such as python -m http.server and uvicorn. This is a standard and expected pattern for the web_server functionality described within the skill.
Evidence: Found in SKILL.md and references/web-endpoints.md for launching various HTTP serving processes.

modal-web-serving