메뉴
Use when creating new skills, editing existing skills, or verifying skills work before deployment
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation. Use when starting creative work that will produce a versioned design spec.
Use when you need to render an lmd-* skill phase or companion, or when a render call fails and you must decide whether the gateway is broken or the call was misaddressed.
Use when implementing any feature or bugfix, before writing implementation code
Fans out one focused subagent per independent problem domain (disjoint files, no shared state), then conflict-scans and integrates. Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies.
Executes a .lmd.md plan in this session — the main agent runs each task itself, pauses at batch checkpoints for human review, and closes at a whole-branch final gate. No companions. Use when executing an implementation plan inline with human review checkpoints.
Guides branch completion by verifying tests, detecting the workspace environment, presenting exactly 4 (or 3 for detached HEAD) integration options, then executing the choice with provenance-based worktree cleanup. Use when implementation is complete and you must decide how to integrate the work — merge, PR, keep, or discard.
| name | lmd-writing-skills |
| description | Use when creating new skills, editing existing skills, or verifying skills work before deployment |
Writing skills IS Test-Driven Development applied to process documentation. Write the pressure test first. Watch the agent fail without the skill. Write the minimal skill. Watch it comply. Close loopholes. Core principle: If you didn't watch an agent fail without the skill, you don't know if the skill teaches the right thing. Violating the letter of the rules is violating the spirit of the rules.
Body renders one phase at a time — never read it from disk.
Rendering, diagnosis and fallback: see the lmd-rendering-skills skill.
REQUIRED BACKGROUND: lmd-test-driven-development — it defines the RED-GREEN-REFACTOR cycle and the Iron Law this skill adapts to documentation.
NO SKILL WITHOUT A FAILING TEST FIRST
Applies to NEW skills AND EDITS. Wrote the skill before the test? Delete it. Start over. Delete means delete.
red → green → refactor. The rationalizations phase is read when tempted to skip a step.
skill-anatomy — what a skill is, types, directory/file structure, SKILL.md template, anti-patternsskill-discovery-optimization — description/keyword/naming/token rules + discovery workflowbulletproofing — close loopholes, rationalization tables, match-the-form-to-the-failuretesting/methodology — RED→GREEN→REFACTOR testing workflow (pressure scenarios, rationalization tables)testing/skill-types — how to test discipline/technique/pattern/reference skillstesting/creation-checklist — TDD (test-driven development)-adapted checklist before deploying a skillclaude-md-testing-example — worked example of a test campaignflowchart-conventions — when to use graphviz, shape/naming conventionsanthropic-best-practices — Anthropic's official skill authoring guidancepersuasion-principles — research foundation for bulletproofingNew or edited skill -> baseline test exists and failed first
Otherwise -> not done
No exceptions without your human partner's permission.