The Agent Skills Directory

[Malicious URL Detection] (CRITICAL): Automated scanning (URLite) identified a phishing domain (Layer.su) within the skill's components, representing a high-severity threat.
[Command Execution] (MEDIUM): File references/platform/command.md provides detailed methods for executing arbitrary shell commands through @effect/platform/Command, allowing an agent to run system-level operations like 'ls' and 'cat'.
[Data Exfiltration] (MEDIUM): The skill demonstrates the ability to read sensitive local files (FileSystem.readFileString) and execute HTTP requests (HttpClient.fetch), creating a functional path for exfiltrating local data to remote endpoints.
[Indirect Prompt Injection] (LOW): The skill defines ingestion points for external API data (GET /todos in references/platform/http.md). Ingestion points: External fetch calls to api.example.demo. Boundary markers: None present in the code examples. Capability inventory: Command execution and file system write operations. Sanitization: No evidence of data sanitization or validation of external content.
[External Downloads] (LOW): The skill relies on external Node.js dependencies from the @effect and @opentelemetry ecosystems, which increase the attack surface.

effect-ts