What problem does it solve? Turning a rough idea into a well-structured AI agent skill is hard: prompts tend to bloat with unnecessary scaffolding, lack eval coverage, and drift from organizational standards. This Skill guides the creation, editing, and quality analysis of SKILL.md-based agent skills through a goal-driven build loop with deterministic linting and token budgets. ## Core Features & Use Cases - Build and Edit Flow: Scaffolds new skills from a template, hardens vague ideas through structured questioning, captures decisions in a memlog, and applies an earn-its-place leanness test to every line. - Analyze Mode: Runs deterministic pre-pass scanners (prompt metrics, workflow integrity) plus LLM lenses over a skill and renders a self-contained HTML or Markdown findings report. - Customization Surface: Emits an optional customize.toml with three-layer merge rules (base, team, personal overrides) for activation hooks, persistent facts, and output paths. - Use Case: A platform engineer wants to codify a runbook as an agent skill. They invoke the builder, which pressure-tests the idea, scaffolds the skill with init_skill.py, runs it on real input, lints it, and ships it under the token budget. ## Quick Start Ask the agent to build a new skill from your idea, or point it at an existing skill directory and ask it to analyze or edit that skill.
