The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data by reading existing documentation files from directories like docs/ and site/ to match neighboring styles and components.
Ingestion points: Files located in docs/, documentation/, content/docs/, site/, and website/docs/ are read for structure and tone.
Boundary markers: The skill contains explicit instructions to ignore bad habits found in neighboring files and enforces two 'Hard Gates' (Persona and Tone gates) that require explicit human approval before content is written.
Capability inventory: The skill has the ability to write new documentation files and read auxiliary instruction files (SKILL.md) from the local file system.
Sanitization: Enforces strict stylistic filters via the simple-english and i-have-adhd sub-skills, which act as a structural and linguistic validation layer.
[DYNAMIC_EXECUTION]: The skill utilizes a modular architecture that requires loading additional instructions from local paths such as .claude/skills/simple-english/SKILL.md and .grok/skills/i-have-adhd/SKILL.md. This dynamic loading is restricted to local configuration files and includes error handling that halts execution if the files are missing.
[COMMAND_EXECUTION]: The skill contains markdown templates for package manager installation commands (npm, yarn, pnpm). These are defined as static documentation snippets intended for output to the user and not for execution by the agent on the host system.

docs