Writing Skill

Overview

The Writing Skill applies Test-Driven Development (TDD) to process documentation.

Personal skills are stored in agent-specific directories (
~/.claude/skills
for Claude Code,
~/.agents/skills/
for Codex)

You write test cases (stress scenarios for sub-agents), observe them fail (baseline behavior), write the skill (documentation), observe the tests pass (agent compliance), then refactor (close loopholes).

Core Principle: If you don't see the agent fail without the skill, you don't know if the skill teaches the right thing.

Required Background: Before using this skill, you must understand the superpower: Test-Driven Development. That skill defines the basic red-green-refactor cycle. This skill adapts TDD for documentation.

Official Guidance: For Anthropic's official skill creation best practices, refer to anthropic-best-practices.md. This document provides additional patterns and guidelines to complement the TDD-centered approach in that skill.

What is a Skill?

A Skill is a reference guide for validated techniques, patterns, or tools. Skills help future Claude instances find and apply effective methods.

Skills are: Reusable techniques, patterns, tools, reference guides

Skills are NOT: Narratives about how you solved a problem once

TDD to Skill Creation Mapping

TDD Concept	Skill Creation
Test Case	Stress scenario with sub-agent
Production Code	Skill documentation (SKILL.md)
Test Failure (Red)	Agent violates rules without the skill (baseline)
Test Pass (Green)	Agent complies with the existing skill
Refactor	Close loopholes while maintaining compliance
Write Test First	Run baseline scenarios before writing the skill
Watch It Fail	Record the exact rationalizations the agent uses
Minimal Code	Write the skill to address these specific violations
Watch It Pass	Verify the agent now complies
Refactor Cycle	Look for new rationalizations → block → re-verify

The entire skill creation process follows red-green-refactor.

When to Create a Skill

Create when:

The technique isn't intuitively obvious to you
You would reference it again in a project
The pattern applies broadly (not project-specific)
Others would benefit

Don't create for:

One-off solutions
Standard practices already well-documented elsewhere
Project-specific conventions (put in CLAUDE.md)
Mechanical constraints (automate if you can enforce via regex/validation - save documentation for judgment calls)

Skill Types

Technique

Specific methods and steps (condition-based waiting, root cause tracing)

Pattern

Ways of thinking about problems (flatten with flags, test invariants)

Reference

API docs, syntax guides, tool documentation (office docs)

Directory Structure

skills/
  skill-name/
    SKILL.md              # Main reference (required)
    supporting-file.*     # Only if needed

Flat Namespace - All skills live in a single searchable namespace

Separate Files:

Heavy Reference (100+ lines) - API docs, comprehensive syntax
Reusable Tools - Scripts, utilities, templates

Keep Inline:

Principles and concepts
Code patterns (<50 lines)
Everything else

SKILL.md Structure

Front Matter (YAML):

Only two fields supported:
```
name
```
and
```
description
```
Maximum 1024 characters total
```
name
```
: Use only letters, numbers, and hyphens (no brackets, special characters)
```
description
```
: Third person, only describes when to use (not what it does)
- Start with "Use when...", focus on trigger conditions
- Include specific symptoms, situations, and context
- Never summarize the skill's process or workflow (see CSO section for why)
- Keep under 500 characters when possible

markdown

---
name: Skill-Name-With-Hyphens
description: Use when [specific triggering conditions and symptoms]
---

# Skill Name

## Overview
What is this? Core principle in 1-2 sentences.

## When to Use
[Small inline flowchart IF decision non-obvious]

Bullet list with SYMPTOMS and use cases
When NOT to use

## Core Pattern (for techniques/patterns)
Before/after code comparison

## Quick Reference
Table or bullets for scanning common operations

## Implementation
Inline code for simple patterns
Link to file for heavy reference or reusable tools

## Common Mistakes
What goes wrong + fixes

## Real-World Impact (optional)
Concrete results

Claude Search Optimization (CSO)

Key to Discovery: Future Claude instances need to find your skill

1. Rich Description Field

Purpose: Claude reads descriptions to decide which skills to load for a given task. Make it answer: "Should I read this skill right now?"

Format: Start with "Use when...", focus on trigger conditions

Key: Description = When to use, not what the skill does

The description should only describe trigger conditions. Do NOT summarize the skill's process or workflow in the description.

Why This Matters: Testing shows that when descriptions summarize the workflow, Claude may follow the description instead of reading the full skill content. Even though the skill's flowchart clearly showed two reviews (spec compliance then code quality), a description of "code review between tasks" led Claude to perform only one review.

When the description was changed to "Use when executing implementation plans with independent tasks in the current session" (no workflow summary), Claude correctly read the flowchart and followed the two-stage review process.

Pitfalls: Descriptions that summarize workflows create shortcuts Claude will take. The skill body becomes documentation Claude skips.

yaml

# ❌ BAD: Summarizes workflow - Claude may follow this instead of reading skill
description: Use when executing plans - dispatches subagent per task with code review between tasks

# ❌ BAD: Too much process detail
description: Use for TDD - write test first, watch it fail, write minimal code, refactor

# ✅ GOOD: Just triggering conditions, no workflow summary
description: Use when executing implementation plans with independent tasks in the current session

# ✅ GOOD: Triggering conditions only
description: Use when implementing any feature or bugfix, before writing implementation code

Content:

Use specific triggers, symptoms, and situations that show the skill applies
Describe the problem (race conditions, inconsistent behavior) not language-specific symptoms (setTimeout, sleep)
Keep triggers technology-agnostic unless the skill itself is technology-specific
If the skill is technology-specific, explicitly state that in the trigger
Write in third person (for injection into system prompts)
Never summarize the skill's process or workflow

yaml

# ❌ BAD: Too abstract, vague, doesn't include when to use
description: For async testing

# ❌ BAD: First person
description: I can help you with async tests when they're flaky

# ❌ BAD: Mentions technology but skill isn't specific to it
description: Use when tests use setTimeout/sleep and are flaky

# ✅ GOOD: Starts with "Use when", describes problem, no workflow
description: Use when tests have race conditions, timing dependencies, or pass/fail inconsistently

# ✅ GOOD: Technology-specific skill with explicit trigger
description: Use when using React Router and handling authentication redirects

2. Keyword Coverage

Use words Claude will search for:

Error messages: "hook timeout", "ENOTEMPTY", "race condition"
Symptoms: "flaky", "hanging", "zombie", "pollution"
Synonyms: "timeout/hang/freeze", "cleanup/teardown/afterEach"
Tools: Actual commands, library names, file types

3. Descriptive Naming

Use active voice, verb first:

✅
```
creating-skills
```
not
```
skill-creation
```

✅

condition-based-waiting

not

async-test-helpers

4. Token Efficiency (Critical)

Problem: Onboarding and frequently referenced skills are included in every conversation. Every token counts.

Target Word Counts:

Onboarding workflows: <150 words each
Frequently used skills: <200 total words
Other skills: <500 words (still be concise)

Techniques:

Move Details to Tool Help:

bash

# ❌ BAD: Document all flags in SKILL.md
search-conversations supports --text, --both, --after DATE, --before DATE, --limit N

# ✅ GOOD: Reference --help
search-conversations supports multiple modes and filters. Run --help for details.

Use Cross-References:

markdown

# ❌ BAD: Repeat workflow details
When searching, dispatch subagent with template...
[20 lines of repeated instructions]

# ✅ GOOD: Reference other skill
Always use subagents (50-100x context savings). REQUIRED: Use [other-skill-name] for workflow.

Compress Examples:

markdown

# ❌ BAD: Verbose example (42 words)
your human partner: "How did we handle authentication errors in React Router before?"
You: I'll search past conversations for React Router authentication patterns.
[Dispatch subagent with search query: "React Router authentication error handling 401"]

# ✅ GOOD: Minimal example (20 words)
Partner: "How did we handle auth errors in React Router?"
You: Searching...
[Dispatch subagent → synthesis]

Eliminate Redundancy:

Don't repeat content from cross-referenced skills
Don't explain what's obvious from the command
Don't include multiple examples of the same pattern

Verification:

bash

wc -w skills/path/SKILL.md
# getting-started workflows: aim for <150 each
# Other frequently-loaded: aim for <200 total

Name by what you do or core insight:

✅

condition-based-waiting

async-test-helpers

✅
```
using-skills
```
not
```
skill-usage
```

✅

flatten-with-flags

data-structure-refactoring

✅
```
root-cause-tracing
```
>
```
debugging-techniques
```

Gerunds (-ing) work for processes:

creating-skills

testing-skills

debugging-with-logs

Active, describes the action you're taking

4. Cross-Reference Other Skills

When writing docs that reference other skills:

Only use the skill name with explicit requirement markers:

✅ GOOD:

**REQUIRED SUB-SKILL:** Use superpowers:test-driven-development

✅ GOOD:

**REQUIRED BACKGROUND:** You MUST understand superpowers:systematic-debugging

❌ BAD:

See skills/testing/test-driven-development

(unclear if required)

❌ BAD:

@skills/testing/test-driven-development/SKILL.md

(heavy cognitive load, burns context)

Why no @ links: The

syntax forces immediate loading of the document, consuming 200k+ context before it's needed.

Flowchart Usage

dot

digraph when_flowchart {
    "Need to show information?" [shape=diamond];
    "Decision where I might go wrong?" [shape=diamond];
    "Use markdown" [shape=box];
    "Small inline flowchart" [shape=box];

    "Need to show information?" -> "Decision where I might go wrong?" [label="yes"];
    "Decision where I might go wrong?" -> "Small inline flowchart" [label="yes"];
    "Decision where I might go wrong?" -> "Use markdown" [label="no"];
}

Flowcharts are only for:

Non-obvious decision points
Process loops where you might stop early
"When to use A vs B" decisions

Never use flowcharts for:

Reference material → tables, lists
Code examples → Markdown blocks
Linear instructions → numbered lists
Non-semantic labels (step1, helper2)

For Graphviz style rules, see @graphviz-conventions.dot.

Visualize for your human partner: Use

render-graphs.js

in this directory to render skill flowcharts as SVG:

bash

./render-graphs.js ../some-skill           # Each diagram separately
./render-graphs.js ../some-skill --combine # All diagrams in one SVG

Code Examples

One excellent example beats many mediocre examples

Choose the most relevant language:

Testing techniques → TypeScript/JavaScript
System debugging → Shell/Python
Data processing → Python

Good Examples:

Complete and runnable
Well-commented, explains why
From real scenarios
Clearly shows the pattern
Ready to adapt (not generic templates)

Don't:

Implement in 5+ languages
Create fill-in-the-blank templates
Write contrived examples

You're good at porting - one great example is enough.

Document Organization

Self-Contained Skill

defense-in-depth/
  SKILL.md    # Everything inline

When: Everything fits, no heavy reference needed

Skill with Reusable Tools

condition-based-waiting/
  SKILL.md    # Overview + patterns
  example.ts  # Working helpers to adapt

When: Tools are reusable code, not just narrative

Skill with Heavy Reference

pptx/
  SKILL.md       # Overview + workflows
  pptxgenjs.md   # 600 lines API reference
  ooxml.md       # 500 lines XML structure
  scripts/       # Executable tools

When: Reference material is too large for inline

Iron Rule (Same as TDD)

NO SKILL WITHOUT A FAILING TEST FIRST

This applies to new skills and edits to existing skills.

Wrote the skill before testing? Delete it. Start over. Edited a skill without testing? Same violation.

No exceptions:

Not for "simple additions"
Not for "just adding a section"
Not for "documentation updates"
Don't keep untested changes as "reference"
Don't "adapt" while running tests
Delete means delete

Required Background: The Superpowers: Test-Driven Development skill explains why this matters. The same principles apply to documentation.

Testing All Skill Types

Different skill types require different testing approaches:

Discipline Enforcement Skills (Rules/Requirements)

Examples: TDD, validate before finish, design before coding

Test with:

Academic questions: Do they understand the rules?
Stress scenarios: Do they comply under pressure?
Multiple stress叠加: Time + sunk cost + fatigue
Identify rationalizations and add explicit counters

Success Criteria: Agent follows rules under maximum pressure

Technique Skills (How-To Guides)

Examples: Condition-based waiting, root cause tracing, defensive programming

Test with:

Application scenarios: Can they apply the technique correctly?
Variation scenarios: Do they handle edge cases?
Missing information tests: Do they indicate gaps?

Success Criteria: Agent successfully applies the technique to new scenarios

Pattern Skills (Mental Models)

Examples: Reduce complexity, information hiding concepts

Test with:

Recognition scenarios: Can they identify when the pattern applies?
Application scenarios: Can they use the mental model?
Counterexamples: Do they know when not to apply it?

Success Criteria: Agent correctly identifies when/how to apply the pattern

Reference Skills (Documentation/API)

Examples: API docs, command references, library guides

Test with:

Retrieval scenarios: Can they find the correct information?
Application scenarios: Can they correctly use what they found?
Gap testing: Are common use cases covered?

Success Criteria: Agent finds and correctly applies reference information

Common Excuses for Skipping Tests

Excuse	Reality
"The skill is obviously clear"	Clear to you ≠ clear to other agents. Test it.
"It's just a reference"	References can have gaps, unclear sections. Test retrieval.
"Testing is overkill"	Untested skills have problems. Always. A 15-minute test saves hours.
"I'll test if there's a problem"	Problems = agents can't use the skill. Test before deployment.
"Testing is too tedious"	Testing is simpler than debugging bad skills in production.
"I'm confident it's good"	Overconfidence causes problems. Test anyway.
"Academic review is enough"	Reading ≠ using. Test application scenarios.
"No time to test"	Deploying untested skills wastes more time fixing it later.

All of this means: Test before deployment. No exceptions.

Bulletproofing Skills Against Rationalization

Discipline-enforcing skills like TDD need to resist rationalization. Agents are smart and will find loopholes under pressure.

Psychology Note: Understanding why persuasion techniques work helps apply them systematically. For research on authority, commitment, scarcity, social proof, and unity principles (Cialdini, 2021; Meincke et al., 2025), see persuasion-principles.md.

Explicitly Block Every Loophole

Don't just state the rule - prohibit specific workarounds:

<bad> ```markdown Write code before test? Delete it. ``` </bad> <good> ```markdown Write code before test? Delete it. Start over.

No exceptions:

Don't keep it as "reference"
Don't "adapt" it while writing tests
Don't look at it
Delete means delete

</good>

### Address "Spirit vs Letter" Arguments

Add a core principle early:

```markdown
**Violating the letter of the rules is violating the spirit of the rules.**

This cuts off an entire class of "I followed the spirit" rationalizations.

Build a Rationalization Table

Capture rationalizations from baseline tests (see Testing section below). Every excuse the agent makes goes in the table:

markdown

| Excuse | Reality |
|--------|---------|
| "Too simple to test" | Simple code breaks. Test takes 30 seconds. |
| "I'll test after" | Tests passing immediately prove nothing. |
| "Tests after achieve same goals" | Tests-after = "what does this do?" Tests-first = "what should this do?" |

Build a Red Flags List

Give agents an easy self-check for rationalization:

markdown

## Red Flags - STOP and Start Over

- Code before test
- "I already manually tested it"
- "Tests after achieve the same purpose"
- "It's about spirit not ritual"
- "This is different because..."

**All of these mean: Delete code. Start over with TDD.**

Update CSO with Violation Symptoms

Add to description: Symptoms that you're about to violate the rule:

yaml

description: use when implementing any feature or bugfix, before writing implementation code

Skill Red-Green-Refactor

Follow the TDD cycle:

Red: Write Failing Test (Baseline)

Run stress scenarios with sub-agents without the skill. Record exact behavior:

What choices did they make?
What rationalizations did they use (verbatim)?
Which pressures triggered violations?

This is "watch the test fail" - you must understand what the agent would naturally do before writing the skill.

Green: Write Minimal Skill

Write the skill to address these specific rationalizations. Don't add extra content for hypothetical cases.

Run the same scenarios with the skill. The agent should now comply.

Refactor: Close Loopholes

Agent found new rationalizations? Add explicit counters. Retest until bulletproof.

Testing Methods: See @testing-skills-with-subagents.md for complete testing methodology:

How to write stress scenarios
Types of pressure (time, sunk cost, authority, fatigue)
Systematically closing loopholes
Meta-testing techniques

Anti-Patterns

❌ Narrative Examples

"In the October 3, 2025 session, we discovered empty projectDir caused..." Why bad: Too specific, not reusable

❌ Multi-Language Dilution

example-js.js, example-py.py, example-go.go Why bad: Mediocre quality, high maintenance burden

❌ Code in Flowcharts

dot

step1 [label="import fs"];
step2 [label="read file"];

Why bad: Can't copy-paste, hard to read

❌ Generic Labels

Helper 1, Helper 2, Step 3, Pattern 4 Why bad: Labels should be semantically meaningful

STOP: Before Moving to Next Skill

After writing any skill, you must stop and complete the deployment process.

Don't:

Batch-create multiple skills without testing each one
Move to the next skill before validating the current one
Skip testing because "batching is more efficient"

The deployment checklist below is mandatory for every skill.

Deploying untested skills = deploying untested code. It's a violation of quality standards.

Skill Creation Checklist (TDD Adapted)

Important: Use TodoWrite to create todos for each checklist item below.

Red Phase - Write Failing Test:

Create stress scenarios (3+ combined pressures for discipline skills)
Run scenarios without the skill - record baseline behavior verbatim
Identify rationalization/failure patterns

Green Phase - Write Minimal Skill:

Name uses only letters, numbers, hyphens (no brackets/special characters)
YAML frontmatter only includes name and description (max 1024 characters)
Description starts with "Use when" and includes specific triggers/symptoms
Description written in third person
Search keywords included (errors, symptoms, tools)
Clear overview with core principle
Addresses specific baseline failures identified in RED phase
Code is inline or linked to separate files
One excellent example (not multi-language)
Run scenarios with the skill - verify agent now complies

Refactor Phase - Close Loopholes:

Identify new rationalizations from testing
Add explicit counters (if discipline skill)
Build rationalization table from all test iterations
Create red flags list
Retest until bulletproof

Quality Check:

Only use small flowcharts when decisions are non-obvious
Quick reference table included
Common mistakes section included
No narrative stories
Supporting files only for tools or heavy reference

Deployment:

Commit skill to git and push to your fork (if configured)
Consider contributing via PR if broadly useful

Discovery Workflow

How future Claude instances find your skill:

Encounter problem ("tests are flaky")
Find skill (description matches)
Scan overview (is this relevant?)
Read pattern (quick reference table)
Load example (only when implementing)

Optimize for this workflow - place searchable terms early and often.

Bottom Line

Skill creation is TDD for process documentation.

Same iron rule: No skill without a failing test first. Same cycle: Red (baseline) → Green (write skill) → Refactor (close loopholes). Same benefits: Better quality, fewer surprises, bulletproof results.

If you follow TDD for code, follow TDD for skills. It's the same rules applied to documentation.

writing-skills

NPX Install

Tags

SKILL.md Content (Chinese)

Writing Skill

Overview

What is a Skill?

TDD to Skill Creation Mapping

When to Create a Skill

Skill Types

Technique

Pattern

Reference

Directory Structure

SKILL.md Structure

Claude Search Optimization (CSO)

1. Rich Description Field

2. Keyword Coverage

3. Descriptive Naming

4. Token Efficiency (Critical)

4. Cross-Reference Other Skills

Flowchart Usage

Code Examples

Document Organization

Self-Contained Skill

Skill with Reusable Tools

Skill with Heavy Reference

Iron Rule (Same as TDD)

Testing All Skill Types

Discipline Enforcement Skills (Rules/Requirements)

Technique Skills (How-To Guides)

Pattern Skills (Mental Models)

Reference Skills (Documentation/API)

Common Excuses for Skipping Tests

Bulletproofing Skills Against Rationalization

Explicitly Block Every Loophole

Build a Rationalization Table

Build a Red Flags List

Update CSO with Violation Symptoms

Skill Red-Green-Refactor

Red: Write Failing Test (Baseline)

Green: Write Minimal Skill

Refactor: Close Loopholes

Anti-Patterns

❌ Narrative Examples

❌ Multi-Language Dilution

❌ Code in Flowcharts

❌ Generic Labels

STOP: Before Moving to Next Skill

Skill Creation Checklist (TDD Adapted)

Discovery Workflow