The Agent Skills Directory

[Data Exposure & Exfiltration] (LOW): The skill sends infrastructure metadata and generated diagram code to https://app.eraser.io. This domain is not part of the trusted whitelist. Users should ensure they trust the third-party service with their architecture details.
[Command Execution] (LOW): The skill uses curl to make API requests. While the endpoint is static, the payload is dynamically generated from user-provided data.
[Prompt Injection] (LOW): The skill processes untrusted CloudFormation and CLI output files. There is a risk that malicious content within these files could attempt to manipulate the agent's output or actions. (1) Ingestion points: SKILL.md describes parsing CloudFormation templates and AWS CLI JSON output provided by the user. (2) Boundary markers: None identified. The instructions do not tell the agent to ignore instructions embedded in the source data. (3) Capability inventory: The skill uses Bash(curl) to communicate with an external API. (4) Sanitization: No specific sanitization or validation of the input data is mentioned before it is processed into DSL.

aws-diagrams