The Agent Skills Directory

[SAFE]: The skill is purely informational and defines a workflow for the agent to follow. It provides guidance on how to search for existing solutions in package registries (npm, PyPI), GitHub, and MCP servers.
[DATA_EXPOSURE]: The skill mentions checking local configuration files such as ~/.claude/settings.json and ~/.claude/skills/. This is described as a search step for the agent to see what tools are already available in its environment, which is standard behavior for development-oriented agents and does not involve exfiltration.
[COMMAND_EXECUTION]: While the skill mentions commands like npm install and rg (ripgrep) in examples and decision matrices, these are presented as recommended actions for the agent to take as part of a development task, rather than hidden or malicious payloads.

search-first