Writing Skills

Overview

Create effective, well-tested Claude Code skills that trigger reliably, load efficiently, and produce consistent results. This skill enforces a TDD approach to skill creation — define test prompts first, write the minimal skill, then harden against misfire and rationalization.

This is a RIGID skill. Every step must be followed exactly. No shortcuts.

Phase 1: Define Test Prompts (RED)

[HARD-GATE] Before writing any skill content, define the prompts that should trigger it.

Write 3-5 test prompts:

#	Test Type	Purpose	Example
1	Obvious trigger	Should clearly match	"Create a new React component with tests"
2	Subtle trigger	Should also match	"I need a reusable button"
3	Near-miss	Should NOT match	"Review this component code"
4	Edge case	Define expected behavior	"Create a component and write its docs"
5	Ambiguous	Define which skill wins	"Set up the frontend"

Document expected behavior for each:

Test prompt: "Create a new React component with tests"
Expected: Should trigger component-creation skill
Should NOT trigger: testing-strategy, code-review, writing-skills

STOP — Do NOT proceed to Phase 2 until all test prompts are documented.

Phase 2: Write Minimal SKILL.md (GREEN)

Create the minimum skill definition that passes all test prompts from Phase 1.

Required Structure

Every SKILL.md MUST contain ALL of these elements:

Element	Required	Purpose
Frontmatter (name, description)	Yes	Identification and trigger matching
CSO-optimized description	Yes	Starts with "Use when...", lists trigger conditions
Overview (2-3 sentences)	Yes	Purpose and value
Multi-phase process with STOP markers	Yes	Numbered phases with clear actions
Decision tables	Yes	Tables for choosing between approaches
Anti-patterns / Common mistakes table	Yes	"What NOT to do" section
Anti-rationalization guards	Yes	HARD-GATE or "Do NOT skip" markers
Integration points table	Yes	How skill connects to other skills
Concrete examples	Yes	Code/command examples where non-obvious
Skill type declaration	Yes	RIGID or FLEXIBLE at the end

Frontmatter Rules

---
name: skill-name          # lowercase with hyphens, unique identifier
description: "Use when [trigger conditions]"  # Max 1024 chars
---

[HARD-GATE] The description field MUST start with "Use when".

CSO (Claude Search Optimization) Rules

The description field determines when a skill is selected. Optimize it like a search query.

Rule	Explanation
Start with "Use when..."	Trigger condition format
List specific conditions, not capabilities	What activates the skill, not what it teaches
Maximum 1024 characters	Be concise
Use action verbs	"creating", "debugging", "designing", "reviewing"

Good:

description: "Use when creating database migrations, designing table schemas, adding indexes, or optimizing SQL queries"

Bad:

description: "A comprehensive guide to database design covering normalization, indexing, query optimization, and migration strategies"

Trigger Condition Patterns

Pattern	Format	Example
Action-based	"Use when creating..., updating..., deleting..."	CRUD operations
Problem-based	"Use when debugging..., fixing..., resolving..."	Bug fixes
Artifact-based	"Use when working with Docker files, CI configs..."	File types
Phase-based	"Use when starting a project, finishing a feature..."	Workflow stage

STOP — Verify the description would match test prompts from Phase 1 before proceeding.

Phase 3: Harden the Skill (REFACTOR)

Review and close loopholes:

Check	Question	Fix If Failing
Over-triggering	Does it match prompts it should NOT?	Narrow the description
Under-triggering	Does it miss valid prompts?	Add missing trigger conditions
Rationalization	Would an agent find a way to skip steps?	Add "Do NOT skip" constraints
Ambiguity	Are instructions interpretable multiple ways?	Make them concrete with examples
Token bloat	Is it over 500 lines?	Move reference material to separate files
Missing gates	Are there checkpoints between phases?	Add STOP markers

Token Efficiency Targets

Skill Type	Target Lines	Strategy
Getting-started workflows	< 100 lines	Minimal steps
Frequently-loaded skills	< 200 lines	Tables over prose
Comprehensive reference skills	< 400 lines	Split into phases
Maximum allowed	500 lines	Move extras to separate files

Token Reduction Strategies

• Use tables instead of prose for structured information
• Use terse imperative sentences, not explanatory paragraphs
• Move reference material to separate files (loaded only when needed)
• Eliminate redundancy — say it once
• Use code examples only when the pattern is non-obvious

STOP — Re-run test prompts mentally. Does the skill trigger correctly for all 5? Does it NOT trigger for near-misses?

Phase 4: Validate and Save

1. Verify all 10 required elements from Phase 2 are present
2. Verify token budget is within target
3. Verify STOP markers exist between phases
4. Save the SKILL.md file
5. Test by invoking the skill with each test prompt

Validation Checklist

#	Check	Status
1	Frontmatter has name and description	[ ]
2	Description starts with "Use when"	[ ]
3	Overview is 2-3 sentences	[ ]
4	Phases are numbered with STOP markers	[ ]
5	Decision tables present (not prose)	[ ]
6	Anti-patterns table present	[ ]
7	Anti-rationalization guards present	[ ]
8	Integration points table present	[ ]
9	Concrete examples present	[ ]
10	Skill type declared (RIGID/FLEXIBLE)	[ ]
11	Under 500 lines total	[ ]

Skill Types and Testing

Testing by Skill Type

Skill Type	Test Method	What to Verify
Technique (TDD, debugging)	Apply to problem X and unusual problem Y	Produces right steps; adapts correctly
Pattern (design patterns)	Show code X (matches) and code Y (near-miss)	Identifies correctly; rejects correctly
Reference (API docs, checklists)	Ask "what is the rule for X?" and "what about edge case Z?"	Finds right answer; handles gaps
Discipline (security review)	Review correct code, then review under time pressure	Passes clean code; enforces all rules even under pressure

Anti-Patterns / Common Mistakes

Mistake	Why It Is Wrong	What To Do Instead
Description describes content, not triggers	Skill never gets selected	Start with "Use when..." and list trigger conditions
Too broad — triggers on everything	Useful for nothing	Narrow to specific actions and artifacts
Too long — 1000+ lines	Wastes context tokens every invocation	Stay under 500 lines; split reference material
Missing constraints	Agents rationalize skipping steps	Add "Do NOT" rules and HARD-GATE markers
Untested against real prompts	Will misfire in production	Write test prompts BEFORE writing the skill
External file dependencies	Fragile — files may not exist	Self-contained or check file existence first
Prose where tables work	Harder to scan, more tokens	Use tables for structured comparisons
No STOP markers between phases	Agent blasts through without checkpoints	Add STOP after each phase

Anti-Rationalization Guards

• [HARD-GATE] Do NOT write a skill without defining test prompts first (Phase 1)
• [HARD-GATE] Do NOT omit the description field or use a non-"Use when" format
• Do NOT skip the hardening phase — every skill needs rationalization review
• Do NOT skip the validation checklist — all 11 items must pass
• Do NOT create skills over 500 lines — split into skill + reference file
• Do NOT use prose where a table would be clearer

Integration Points

Skill	Relationship
self-learning	Skills are discovered and loaded via the skill system
auto-improvement	Tracks skill effectiveness and suggests improvements
spec-writing	Skills can be specified using JTBD methodology
code-review	Skills should be reviewed for quality before deployment
testing-strategy	Test prompts follow testing methodology

Concrete Example: Complete Minimal Skill

---
name: docker-setup
description: "Use when creating Dockerfiles, docker-compose configurations, or containerizing an application for development or production"
---

# Docker Setup

## Overview

Guide the creation of Docker configurations for development and production environments. Produces Dockerfiles, docker-compose files, and .dockerignore with best practices.

## Phase 1: Discovery
[Ask about: runtime, ports, volumes, env vars, multi-stage needs]
STOP — confirm requirements.

## Phase 2: Generate Configuration
[Create Dockerfile, docker-compose.yml, .dockerignore]
STOP — present for review.

## Phase 3: Verify
[Build image, run container, verify health check]

## Anti-Patterns
| Mistake | Fix |
|---------|-----|
| Running as root | Use non-root USER |
| No .dockerignore | Always create one |

## Skill Type
**Flexible**

Skill Type

Rigid — The TDD approach (test prompts first), required structure elements, and validation checklist must be followed exactly. No elements may be skipped.