Skill Invocation Guard

Problem

Agents may skip applicable skills before answering, reading files, or changing code, especially when the task looks simple or needs context first. Missing a skill can bypass workflow constraints such as TDD, debugging process, or platform tool mapping, causing rework later.

How It Works

This skill makes the skill check a mandatory pre-response step: if there is even a 1% chance a skill may apply, invoke it before acting. It sets a priority order:
- User instructions (CLAUDE.md, GEMINI.md, AGENTS.md, or direct requests) win
- Superpowers skills override default system behavior
- Default prompts have the lowest priority

It also flags common rationalizations—such as “the question is simple,” “explore the code first,” or “I remember this skill”—as signals to stop and check. The skill separates rigid workflows that should be followed exactly from flexible patterns that can adapt. When several skills may apply, run process skills first, then implementation skills.

Boundaries

It works in environments that can load skills, such as Claude Code or Gemini CLI, but it does not provide task-specific implementation logic. If the user explicitly disables a workflow, the user instruction prevails; if an invoked skill is not relevant, its content can be ignored.