The Agent Skills Directory

[SAFE]: The skill is primarily informational, providing comprehensive guidance and best practices for creating and managing skills in a development environment.
[COMMAND_EXECUTION]: The instructions include examples of shell commands for testing, debugging, and validation (e.g., yarn lint:skill, yarn workspace @local/claude-hooks run:skill). These are standard development tools and target local workspace scripts rather than arbitrary or malicious code.
[DATA_EXPOSURE]: The skill references framework configuration files such as .claude/skills/skill-rules.json and .claude/hooks/skill-activation-prompt.ts. These references are part of the intended architecture and do not involve sensitive user data, credentials, or private keys.
[INDIRECT_PROMPT_INJECTION]: The skill describes a system for proactive suggestions based on user prompt matching (hooks). While it documents how to trigger agent behavior, the skill itself does not ingest untrusted data or include instructions that would bypass safety guidelines.

writing-skills