using-superpowers

To enforce an AI-assistant workflow, you apply a skill-first gating mechanism that identifies and invokes applicable skills before any user-facing response. This prevents hallucinations and scope creep by loading contextual guardrails first.

Automating skill discovery at the start of a conversation involves running a predefined check that routes tasks appropriately. This ensures the AI loads the correct SKILL.md context and follows a deterministic invocation flow before answering.

You load contextual guardrails by activating skills that require a YAML frontmatter with name and description, plus a Markdown body loaded as context. This ensures safe, accurate responses by enforcing deterministic workflow rules.

You need a Skill unit defined by a YAML frontmatter containing the name and description, followed by a Markdown body. Optional resource directories for scripts, references, and assets can also be attached to support the workflow.

Your AI assistant skips relevant skills and produces misaligned outputs because it lacks a deterministic invocation flow. Implementing a skill-first workflow checks for applicable skills before formulating any answer, reducing scope creep.

Yes, skill invocation gating applies across clarification, planning, and execution tasks. It announces and follows a predefined workflow before answering, ensuring proper context loading and safe execution throughout the conversation.