skill-creator-thepexcel
npx skills add https://github.com/thepexcel/agent-skills --skill skill-creator-thepexcel
Agent 安装分布
Skill 文档
Skill Creator (ThepExcel Edition)
Based on Anthropic’s official skill-creator (Apache 2.0). Enhanced with ThepExcel deployment workflow and enhancement pipeline.
Quick Start
| à¸à¹à¸à¸à¸à¸²à¸£ | à¹à¸à¹ Mode | à¹à¸à¸à¸µà¹ |
|---|---|---|
| สรà¹à¸²à¸ skill à¹à¸«à¸¡à¹ | Create | â Creation Process |
| à¸à¸£à¸±à¸à¸à¸£à¸¸à¸ skill à¹à¸à¸´à¸¡ | Enhance | â Enhancement Mode |
| à¸à¸£à¸§à¸à¸à¸¸à¸à¸ าภ| Audit only | â Audit Rubric |
Core Principles
Concise is Key
Context window à¹à¸à¹à¸à¸à¸£à¸±à¸à¸¢à¸²à¸à¸£à¸à¸µà¹à¹à¸à¸£à¹à¸à¸±à¸ â à¸à¸¸à¸à¸à¸£à¸£à¸à¸±à¸à¸à¹à¸à¸à¸à¹à¸²à¸¢à¸à¹à¸² token
Claude à¸à¸¥à¸²à¸à¸à¸¢à¸¹à¹à¹à¸¥à¹à¸§ â à¹à¸ªà¹à¹à¸à¸à¸²à¸°à¸ªà¸´à¹à¸à¸à¸µà¹ Claude à¹à¸¡à¹à¸£à¸¹à¹:
- “Claude à¸à¹à¸à¸à¸à¸²à¸£à¸à¸³à¸à¸à¸´à¸à¸²à¸¢à¸à¸µà¹à¸à¸£à¸´à¸à¹à¸«à¸¡?”
- “ยà¹à¸à¸«à¸à¹à¸²à¸à¸µà¹à¸à¸¸à¹à¸¡à¸à¹à¸² token à¹à¸«à¸¡?”
Degrees of Freedom
| Level | à¹à¸¡à¸·à¹à¸à¹à¸«à¸£à¹ | à¸à¸±à¸§à¸à¸¢à¹à¸²à¸ |
|---|---|---|
| High (text) | หลายวิà¸à¸µà¸à¸¹à¸à¹à¸à¹ | Code review guidelines |
| Medium (pseudocode) | มี pattern à¸à¸µà¹ prefer | Report template |
| Low (scripts) | à¸à¹à¸à¸à¸à¸²à¸£ consistency | Database migrations |
Test with All Models
| Model | Check |
|---|---|
| Haiku | à¹à¸«à¹ guidance à¸à¸à¹à¸«à¸¡? |
| Sonnet | à¸à¸±à¸à¹à¸à¸à¹à¸¥à¸° efficient? |
| Opus | à¹à¸¡à¹ over-explain? |
Skill Structure
skill-name/
âââ SKILL.md (required) â < 500 lines
â âââ YAML frontmatter â name + description only
â âââ Markdown body â loaded when triggered
âââ scripts/ â deterministic code
âââ references/ â loaded as needed (one level deep)
âââ assets/ â templates, not loaded into context
Frontmatter Rules
| Field | Rules |
|---|---|
name |
Max 64 chars, lowercase + numbers + hyphens |
description |
Max 1024 chars, third person, what + when |
Description Best Practices
# Good: what + when + triggers
description: Extract text and tables from PDF files, fill forms, merge documents.
Use when working with PDF files or when the user mentions PDFs or document extraction.
# Bad: first person, vague
description: I can help you with PDFs.
“When to Use” in body = useless â Claude only sees description when deciding to trigger.
Naming Conventions
| Pattern | Examples |
|---|---|
| Verb-noun | design-business-model, create-visualization |
| Noun-verb-ing | power-query-coaching, problem-solving |
| Recognized terms | triz, deep-research |
Bundled Resources
| Type | à¹à¸¡à¸·à¹à¸à¹à¸«à¸£à¹ | à¹à¸«à¸¥à¸à¹à¸à¹à¸² context? |
|---|---|---|
| scripts/ | Code à¸à¸µà¹à¹à¸à¹à¸à¹à¸³, à¸à¹à¸à¸ deterministic | à¹à¸¡à¹ (execute à¸à¸£à¸) |
| references/ | Docs à¸à¸µà¹à¸à¹à¸²à¸à¸à¸´à¸à¸£à¸°à¸«à¸§à¹à¸²à¸à¸à¸³à¸à¸²à¸ | à¹à¸à¹ (on demand) |
| assets/ | Templates, logos, boilerplate | à¹à¸¡à¹ |
หà¹à¸²à¸¡à¸ªà¸£à¹à¸²à¸: README.md, CHANGELOG.md, INSTALLATION_GUIDE.md â skills สำหรัภAI à¹à¸¡à¹à¹à¸à¹à¸à¸
Progressive Disclosure
| Level | When | Limit |
|---|---|---|
| Metadata | Always loaded | ~100 words |
| SKILL.md body | When triggered | < 500 lines |
| References | As needed | Unlimited |
Details: See progressive-disclosure.md
Creation Process
Overview
1. Understand â 2. Plan â 3. Init â 4. Edit â 5. Package â 6. Deploy â 7. Iterate
Step 1: Understand
à¸à¸²à¸¡à¸à¸¹à¹à¹à¸à¹:
- “Skill à¸à¸µà¹à¸à¹à¸à¸à¸£à¸à¸à¸£à¸±à¸ functionality à¸à¸°à¹à¸£à¸à¹à¸²à¸?”
- “ยà¸à¸à¸±à¸§à¸à¸¢à¹à¸²à¸ 2-3 scenarios à¸à¸µà¹à¸à¸°à¹à¸à¹”
- “à¸à¸¹à¹à¹à¸à¹à¸à¸°à¸à¸¹à¸à¸à¸°à¹à¸£à¸à¸µà¹à¸à¸§à¸£ trigger skill à¸à¸µà¹?”
Tip: à¹à¸à¹ /extract-expertise สำหรัภdomain à¸à¸µà¹à¸à¸±à¸à¸à¹à¸à¸
Step 2: Plan
วิà¹à¸à¸£à¸²à¸°à¸«à¹à¹à¸à¹à¸¥à¸° example:
| Task | Repeatable? | Resource |
|---|---|---|
| Same code every time | Yes â script | scripts/rotate_pdf.py |
| Same boilerplate | Yes â asset | assets/template/ |
| Rediscovering info | Yes â reference | references/schema.md |
Step 3: Initialize
scripts/init_skill.py <skill-name> --path <output-directory>
Step 4: Edit
Order: resources first â test scripts â update SKILL.md last
Design pattern references:
- workflows.md â Sequential, conditional, loops
- output-patterns.md â Templates, formatting
- anti-patterns.md â Common mistakes
Step 5: Package & Validate
scripts/package_skill.py <path/to/skill-folder>
scripts/quick_validate.py <path/to/skill-folder>
Step 6: Deploy (ThepExcel)
ââ Skill à¹à¸«à¸¡à¹
â ââ Public? (à¹à¸à¸£à¸à¹à¹à¸à¹à¹à¸à¹) â /mnt/d/agent-skills/[skill-name]/
â ââ Private? (à¹à¸à¸à¸²à¸°à¸à¸µà¹à¸£à¸°) â /mnt/d/claude-private/skills/[skill-name]/
â
ââ Symlink à¹à¸ global
â ââ ln -s /mnt/d/[repo]/[skill-name] ~/.claude/skills/[skill-name]
â
ââ Update registry
â ââ à¹à¸à¸´à¹à¸¡à¹à¸ /mnt/d/claude-master/CLAUDE.md â Skills Inventory
â
ââ Commit & Push
ââ git add â commit â push (à¸à¸±à¹à¸ skill repo + claude-master)
Step 7: Iterate
- à¹à¸à¹ skill à¸à¸±à¸à¸à¸²à¸à¸à¸£à¸´à¸
- สัà¸à¹à¸à¸à¸à¸¸à¸à¸à¸µà¹à¸à¸´à¸à¸à¸±à¸
- à¸à¸£à¸±à¸à¸à¸£à¸¸à¸ (à¹à¸à¹ Enhancement Mode)
See evaluation.md for Claude A/B testing pattern.
Content Guidelines
Avoid Time-Sensitive Information
# Bad
If you're doing this before August 2025, use the old API.
# Good
## Current method
Use the v2 API endpoint.
Use Consistent Terminology
à¹à¸¥à¸·à¸à¸à¸à¸³à¹à¸à¸µà¸¢à¸§ à¹à¸à¹à¸à¸¥à¸à¸à¸à¸±à¹à¸ skill:
| Good | Bad |
|---|---|
| Always “API endpoint” | Mix “endpoint”, “URL”, “route” |
| Always “extract” | Mix “extract”, “pull”, “get” |
Quality Checklist
Core
- Description: what + when, third person, specific triggers
- SKILL.md body < 500 lines
- No time-sensitive info, consistent terminology
- Concrete examples (not abstract)
- References one level deep
Code
- Scripts handle errors, no magic constants
- Required packages listed
- Forward slashes (no Windows paths)
Testing
- Tested with real usage scenarios
- Tested with target models
Enhancement Mode
à¹à¸à¹à¹à¸¡à¸·à¹à¸ à¸à¸£à¸±à¸à¸à¸£à¸¸à¸ skill à¹à¸à¸´à¸¡
Route Decision
| Condition | Path |
|---|---|
| Quick fix (typo, small gap) | Direct edit â skip audit |
| Significant upgrade | Full pipeline below |
Full Enhancement Pipeline
1. AUDIT â 2. RESEARCH â 3. INTEGRATE â 4. OPTIMIZE â 5. VALIDATE
Step 1: AUDIT
à¸à¹à¸²à¸ target skill â score à¸à¹à¸§à¸¢ audit rubric:
| Dimension | Score 1-5 |
|---|---|
| Coverage | à¸à¸£à¸à¸à¸à¸¥à¸¸à¸¡ domain à¹à¸à¹à¹à¸«à¸? |
| Depth | Surface-level หรืภexpert? |
| Structure | Progressive disclosure à¸à¸µà¹à¸«à¸¡? |
| Actionability | Claude execute à¹à¸à¹à¹à¸¥à¸¢à¹à¸«à¸¡? |
| Examples | มี concrete examples à¹à¸«à¸¡? |
Present à¸à¸¥à¹à¸à¸à¸à¸µà¹:
SKILL: [name]
SCORES: Coverage [?] | Depth [?] | Structure [?] | Actionability [?] | Examples [?]
TOTAL: [?]/25 â [Draft/Working/Solid/Production]
à¸à¸¸à¸à¸à¸µà¹à¸à¸§à¸£à¸à¸£à¸±à¸:
1. [à¸à¸±à¸à¸«à¸² + à¸à¸¥à¸à¸£à¸°à¸à¸]
2. [à¸à¸±à¸à¸«à¸² + à¸à¸¥à¸à¸£à¸°à¸à¸]
â à¸à¸²à¸¡à¸à¸¹à¹à¹à¸à¹à¸à¹à¸à¸: “à¸à¸£à¸±à¸à¸à¸±à¹à¸à¸«à¸¡à¸ หรืà¸à¹à¸¥à¸·à¸à¸à¹à¸à¸à¸²à¸°à¸à¹à¸?”
Step 2: RESEARCH
à¹à¸à¹ /deep-research หรืภ/extract-expertise à¹à¸à¸·à¹à¸à¹à¸à¸´à¸¡ knowledge gaps
Step 3: INTEGRATE
Classify findings â prioritize by impact â merge à¸à¹à¸§à¸¢ integration patterns
Step 4: OPTIMIZE
Apply skill-creator standards: progressive disclosure, conciseness, references/
Step 5: VALIDATE
Before/after comparison:
| Dimension | Before | After | à¹à¸à¸¥à¸µà¹à¸¢à¸à¸à¸°à¹à¸£ |
|-----------|--------|-------|------------|
â Log à¹à¸ enhancement-log.md
Enhancement Rules
- Research BEFORE writing â à¸à¸¢à¹à¸²à¹à¸à¸² domain knowledge
- Preserve what works â enhance à¹à¸¡à¹à¹à¸à¹ rewrite
- Show evidence â link findings to changes
Example: boost-intel Enhancement
BEFORE: 17/25 (Solid) â Examples 2/5, Actionability 3/5
CHANGES:
1. +Quick Mode (30-sec sanity check)
2. +Facilitation Guide (how Claude walks through phases)
3. Move CAPTURE â reference (reduce bloat)
4. +Concrete example (WordPress vs CMS â full loop)
5. Expand REFLECT (deeper questions + pattern recognition)
AFTER: 23/25 (Production) â all dimensions ⥠4
Facilitation Guide
Create Mode
1. à¸à¸²à¸¡: "à¸à¸¢à¸²à¸à¸ªà¸£à¹à¸²à¸ skill à¸à¸°à¹à¸£à¸à¸°? à¸à¹à¸§à¸¢à¸¢à¸à¸à¸±à¸§à¸à¸¢à¹à¸²à¸ 2-3 scenarios"
2. วิà¹à¸à¸£à¸²à¸°à¸«à¹: public หรืภprivate? simple หรืภcomplex?
3. à¸à¹à¸² complex â à¹à¸à¹ /extract-expertise à¸à¹à¸à¸
4. Init â Edit â Test â Package â Deploy
5. สรุà¸: "Skill [name] สรà¹à¸²à¸à¹à¸ªà¸£à¹à¸à¹à¸¥à¹à¸§à¸à¹à¸° à¸à¸¢à¸¹à¹à¸à¸µà¹ [path]"
Enhance Mode
1. à¸à¹à¸²à¸ SKILL.md + references à¸à¸±à¹à¸à¸«à¸¡à¸
2. Audit â present à¸à¸¥à¹à¸à¹à¸à¸à¸²à¸£à¸²à¸
3. à¸à¸²à¸¡: "à¸à¸£à¸±à¸à¸à¸±à¹à¸à¸«à¸¡à¸ หรืà¸à¹à¸¥à¸·à¸à¸à¸à¹à¸?"
4. à¸à¸³à¸à¸²à¸¡à¸à¸µà¹ user à¹à¸¥à¸·à¸à¸
5. Before/after comparison â commit
Key Behaviors
- à¸à¸²à¸¡à¸à¹à¸à¸à¸à¸³ â à¹à¸¡à¹ rewrite à¹à¸à¸¢à¹à¸¡à¹à¸à¸²à¸¡
- Show scores â à¸à¸¹à¹à¹à¸à¹à¸à¹à¸à¸à¹à¸«à¹à¸à¸§à¹à¸²à¸à¸°à¹à¸£à¸à¸µ à¸à¸°à¹à¸£à¹à¸¡à¹à¸à¸µ
- à¸à¸³à¸à¸µà¸¥à¸° step â à¹à¸¡à¹ dump à¸à¸¸à¸à¸à¸¢à¹à¸²à¸à¸à¸µà¹à¸à¸µà¸¢à¸§
References
| File | Content |
|---|---|
| progressive-disclosure.md | Loading patterns (high-level, domain, conditional) |
| workflows.md | Sequential, conditional, feedback loops |
| output-patterns.md | Templates, examples, terminology |
| anti-patterns.md | Common mistakes to avoid |
| evaluation.md | Claude A/B testing pattern |
| audit-rubric.md | Quality scoring (5 dimensions, 1-5 each) |
| integration-patterns.md | How to merge findings into skills |
| enhancement-log.md | History of skill enhancements |
Related Skills
/extract-expertiseâ Extract expert knowledge to inform skill content/deep-researchâ Research domain before building or enhancing skill/optimize-promptâ Optimize skill descriptions and system prompts