The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs environment discovery using standard diagnostic commands such as cat, git worktree list, which gh, and git remote -v to identify project structure, existing configurations, and available CLI tools.
[INDIRECT_PROMPT_INJECTION]: The skill surveys the codebase and ingests data from manifest files and third-party AI instruction files (e.g., .cursorrules, .windsurfrules, .clinerules) to automate project setup.
Ingestion points: Survey of project manifest files and existing AI-related rule files in Phase 2.
Capability inventory: Capability to write new instruction files (SKILL.md, CLAUDE.md) and configure automated shell hooks in .claude/settings.json.
Sanitization: A mandatory human review checkpoint is included in Phase 3, requiring the user to approve the synthesized proposal before any files or hooks are written.
[EXTERNAL_DOWNLOADS]: The skill recommends the installation of official platform plugins from the claude-plugins-official registry. These are well-known official resources within the platform ecosystem.

init-new