The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted subpoenas and investigation notices, creating an opportunity for indirect prompt injection. A malicious file could be designed to manipulate the agent's extraction logic or influence subsequent actions such as tool execution. • Ingestion points: Processes legal document files provided via local paths or user uploads. • Boundary markers: Lacks specific delimiters or instructions to isolate document text from the agent's internal logic. • Capability inventory: Capable of reading and writing local files and calling other specialized agent skills. • Sanitization: Does not implement validation or sanitization of document content before processing.
[COMMAND_EXECUTION]: The skill uses a user-supplied --slug argument to construct file system paths for storing analysis results. Without strict validation of this identifier, it represents a potential path traversal surface that could be used to write files outside of the intended project directory.

subpoena-triage