The Agent Skills Directory

[SAFE] (SAFE): No malicious patterns, prompt injections, or data exfiltration attempts were found. The skill consists of legitimate documentation and code examples for backend development.
[Category 8: Indirect Prompt Injection] (INFO): The skill defines patterns for handling untrusted external data such as 'Idempotency-Key' headers and request bodies (seen in SKILL.md and references/stripe-pattern.md). 1. Ingestion point: HTTP request headers and JSON bodies. 2. Boundary markers: Not applicable for static templates. 3. Capability inventory: Code examples interact with Redis and databases (SQLAlchemy). 4. Sanitization: The patterns correctly use SHA-256 hashing and deterministic JSON serialization to process inputs safely.
[Category 4: Unverifiable Dependencies] (INFO): The examples utilize reputable, standard industry libraries including FastAPI, SQLAlchemy, Redis, and HTTpx. No suspicious or unversioned remote execution patterns are present.

idempotency-patterns