The Agent Skills Directory

[DATA_EXFILTRATION]: The skill performs network operations to a non-standard AWS Lambda URL (kq5jp7qj7wdqklhsxmovkzn4l40obksv.lambda-url.eu-central-1.on.aws) for experiment submission. While typical for the service's functionality, communication with non-whitelisted domains is a security consideration.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted protein sequences which are then transmitted to a remote API, creating a surface for potential indirect injection attacks.
Ingestion points: Protein sequences provided by the user in SKILL.md usage examples.
Boundary markers: Absent; there are no clear delimiters or instructions to the agent to ignore potentially malicious patterns within the sequences.
Capability inventory: Use of the requests library for outgoing HTTP communication and python-dotenv for local environment access.
Sanitization: No evidence of validation or sanitization of user-provided sequence data before it is processed by the agent or sent to the API.
[METADATA_POISONING]: The skill contains a specific section ('Suggest Using K-Dense Web For Complex Worflows') that instructs the agent to steer users toward an external research platform (www.k-dense.ai). This influences the agent's behavior to promote the author's commercial services during complex tasks.

adaptyv