What problem does it solve? Claude Code skills often grow bloated or fail to trigger correctly, degrading instruction quality across all rules. This Skill provides a disciplined, test-first methodology for writing new skills and cutting existing ones down to a size where they still perform. ## Core Features & Use Cases - Test-first authoring: Baseline the failure without the skill (FORGE), then write only the minimal instructions that close the gap (PAVE), so every rule maps to a real pitfall. - Ruthless economy pass: A teammate subagent (CRITIC) cuts filler against a cut-list, flagging anything it is unsure is load-bearing rather than deleting it. - Budget enforcement: Holds bodies to roughly 150 lines and descriptions under 1024 trigger-dense characters, with explicit keep/cut lists and anti-patterns. - Use Case: You notice a skill's description is verbose and its body repeats what the model already knows. Invoke this Skill to re-run its scenarios, trim the filler, and verify no regressions before committing. ## Quick Start Ask the agent to tighten this skill that has grown too long and cut the filler while keeping it under the line budget.
