The Agent Skills Directory

[Data Exposure & Exfiltration] (LOW): The skill performs network operations to a non-whitelisted domain: kq5jp7qj7wdqklhsxmovkzn4l40obksv.lambda-url.eu-central-1.on.aws. This is the documented API endpoint for the service and is expected behavior for its stated purpose.
[Unverifiable Dependencies] (LOW): The skill documentation recommends installing requests and python-dotenv via uv pip. These are standard, high-trust packages in the Python ecosystem.
[Indirect Prompt Injection] (LOW):
Ingestion points: Protein sequences are ingested in FASTA format via the sequences field in SKILL.md and reference/api_reference.md.
Boundary markers: None present for the sequence data.
Capability inventory: Uses requests.post to transmit data to an external API.
Sanitization: No explicit sanitization of input sequences is shown in the examples, which is typical for this domain but represents a theoretical surface for data-driven influence if results are later processed by an LLM.

adaptyv