The Agent Skills Directory

Category 8: Indirect Prompt Injection (LOW): The description_generator.py script interpolates external data from tool_spec into templates to generate markdown descriptions. If this generated content is used directly in an agent's system prompt or tool definitions without further sanitization, it could serve as a vector for indirect prompt injection.
Ingestion points: generate_tool_description and ErrorMessageGenerator.generate accept dictionary contexts.
Boundary markers: None present in the templates to differentiate between template structure and user-provided data.
Capability inventory: The script is limited to string manipulation and does not perform file writes, network requests, or command execution.
Sanitization: The script performs no escaping or validation of the input fields before formatting them into the output strings.
General Security Posture (SAFE): Both scripts/description_generator.py and scripts/verify.py use standard library modules (typing, re, pathlib, sys) and follow best practices for local script execution and data modeling. No network operations, hardcoded credentials, or obfuscated code were detected.

tool-design