The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx to fetch and run @jkeskikangas/skillcheck@0.2.4, a validation tool from the author's official namespace. It also mentions agnix as an optional third-party linter.
[REMOTE_CODE_EXECUTION]: The workflow incorporates npx for running validation tools. It includes specific safety warnings for the third-party tool agnix, requiring explicit user opt-in and version pinning to mitigate risks associated with remote code execution.
[COMMAND_EXECUTION]: The skill is configured with allowed-tools for shell execution, including node, npm, npx, and git. These are used for skill validation, building local packages, and inspecting repository state, which is appropriate for a development-focused skill.
[INDIRECT_PROMPT_INJECTION]: As a tool designed to generate and update agent instructions, the skill handles user requirements that influence downstream agent behavior. This surface is managed through a mandatory generator-critic workflow.
Ingestion points: User-provided prompts for new skills and content from existing SKILL.md files.
Boundary markers: The skill explicitly directs the agent to target a specific rubric and use a fresh-context critic ($reviewing-skills) to audit generated content.
Capability inventory: The skill possesses file write, edit, and command execution capabilities (via node/npm/npx) to create and validate artifacts.
Sanitization: Content is sanitized through automated linters (skillcheck) and qualitative review gates to ensure the resulting skills meet safety and quality bars.
[DYNAMIC_EXECUTION]: The skill includes instructions to perform local builds using npm install and npm run build if pre-compiled binaries for the skillcheck tool are missing in the local environment.

writing-skills