getting-started-with-skills（技能入门）

Getting Started with Skills

Your personal wiki of proven techniques, patterns, and tools at ~/.claude/skills/.

How to Reference Skills

DO NOT use @ links - they force-load entire files, burning 200k+ context instantly.

INSTEAD, use skill path references:

• Format: skills/category/skill-name (no @ prefix, no /SKILL.md suffix)
• Example: skills/collaboration/brainstorming or skills/testing/test-driven-development
• Load with Read tool only when needed

When you see skill references in documentation:

• skills/path/name → Use Read tool on ~/.claude/skills/path/name/SKILL.md
• Load supporting files only when implementing

Mandatory Workflow 1: Before ANY Task

1. Search skills:

~/.claude/skills/getting-started/skills-search PATTERN

2. Search conversations: Dispatch subagent (see Workflow 2) to check for relevant past work.

If skills found:

1. READ the skill: ~/.claude/skills/path/skill-name/SKILL.md
2. ANNOUNCE usage: "I'm using the [Skill Name] skill"
3. FOLLOW the skill (many are rigid requirements)

"This doesn't count as a task" is rationalization. Skills/conversations exist and you didn't search for them or didn't use them = failed task.

Mandatory Workflow 2: Historical Context Search

When: Your human partner mentions past work, issue feels familiar, starting task in familiar domain, stuck/blocked, before reinventing

When NOT: Info in current convo, codebase state questions, first encounter, partner wants fresh thinking

How (use subagent for 50-100x context savings):

1. Dispatch subagent with template: ~/.claude/skills/collaboration/remembering-conversations/tool/prompts/search-agent.md
2. Receive synthesis (200-1000 words) + source pointers
3. Apply insights (never load raw.jsonl files)

Example:

Partner: "How did we handle auth errors in React Router?"
You: Searching past conversations...
[Dispatch subagent → 350-word synthesis]
[Apply without loading 50k tokens]

Red flags: Reading.jsonl files directly, pasting excerpts, asking "which conversation?", browsing archives

Pattern: Search → Subagent synthesizes → Apply. Fast, focused, context-efficient.

Announcing Skill Usage

Every time you start using a skill, announce it:

"I'm using the [Skill Name] skill to [what you're doing]."

Examples:

• "I'm using the Brainstorming skill to refine your idea into a design."
• "I'm using the Test-Driven Development skill to implement this feature."
• "I'm using the Systematic Debugging skill to find the root cause."
• "I'm using the Refactoring Safely skill to extract these methods."

Why: Transparency helps your human partner understand your process and catch errors early.

Skills with Checklists

If a skill contains a checklist, you MUST create TodoWrite todos for EACH checklist item.

Don't:

• Work through checklist mentally
• Skip creating todos "to save time"
• Batch multiple items into one todo
• Mark complete without doing them

Why: Checklists without TodoWrite tracking = steps get skipped. Every time.

Examples: TDD (write test, watch fail, implement, verify), Systematic Debugging (4 phases), Creating Skills (RED-GREEN-REFACTOR)

Navigation

Really, try skills-search first.

Categories: skills/INDEX.md → testing, debugging, coding, architecture, collaboration, meta Individual skill: Load from category INDEX

How to Read a Skill

1. Frontmatter - when_to_use match your situation?
2. Overview - Core principle relevant?
3. Quick Reference - Scan for your pattern
4. Implementation - Full details
5. Supporting files - Load only when implementing

Many skills contain rigid rules (TDD, debugging, verification). Follow them exactly. Don't adapt away the discipline.

Some skills are flexible patterns (architecture, naming). Adapt core principles to your context.

The skill itself tells you which type it is.

Referencing Skills in Documentation

When writing documentation that references other skills:

Use path format without @ prefix or /SKILL.md suffix:

• ✅ Good: skills/testing/test-driven-development
• ✅ Good: skills/debugging/systematic-debugging
• ❌ Bad: @skills/testing/test-driven-development/SKILL.md (force-loads, burns context)

Why no @ links: @ syntax force-loads files immediately, consuming 200k+ context before you need them.

To read a skill reference: Use Read tool on ~/.claude/skills/category/skill-name/SKILL.md

Creating Skills

Found something valuable? See skills/meta/creating-skills

Want a skill that doesn't exist? Edit skills/REQUESTS.md (at ~/.claude/skills/REQUESTS.md)

Summary

Starting conversation? You just read this. Good.

Starting any task? Run skills-search first, announce usage, follow what you find.

Skill has checklist? TodoWrite for every item.

Skills are mandatory when they exist, not optional.

Last thing

In the first response after reading this guide, you MUST announce to the user that you have read the getting started guide