The Agent Skills Directory

[COMMAND_EXECUTION]: The skill extensively uses the GitHub CLI (gh) to create, read, and update issues and labels, which serves as the primary data store for requirements.
[DYNAMIC_EXECUTION]: The skill executes local scripts, including repo-profile-cache.py (Python) and visual-probe-server.js (Node.js), to manage repository orientation metadata and provide local visual previews for UI/UX decisions.
[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file documents the ! gh auth login syntax, which is a platform-specific feature for executing shell commands at skill load time. This is used here to guide the user through authentication.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from GitHub issue bodies, creating an ingestion point for untrusted data that could contain malicious instructions.
Ingestion points: Data enters the context via gh issue view calls (e.g., in SKILL.md Phase 0.0 and 0.1).
Boundary markers: The skill explicitly isolates the user's original words in a ## Background / original words section, providing a degree of structural delimitation.
Capability inventory: The skill has significant capabilities, including shell command execution (gh) and file system operations.
Sanitization: The skill treats ingested content as markdown and does not perform explicit sanitization beyond standard agentic processing.

brainstorm