The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs local file system operations to scaffold tool classes and install base classes. Evidence: SKILL.md instructions specify copying claude_style_tool.py to a project's utils directory and generating new .py files in the tools directory.
[SAFE]: The skill is primarily focused on improving the security posture of agentic tools. It implements a 'fail-closed' design where critical safety properties like is_read_only and is_destructive default to restrictive values if not explicitly set.
[DATA_EXPOSURE]: The skill instructions contain hardcoded local file paths (e.g., /Users/jigege/ai-base/...). While these disclose environment details, they do not expose sensitive credentials or personal data.
[PROMPT_INJECTION]: The skill uses natural language instructions to guide the agent and does not contain attempts to bypass safety filters or override system prompts.

langchain-tool-builder