// Skills wiki intro - mandatory workflows, search tool, brainstorming triggers
| name | Getting Started with Skills |
| description | Skills wiki intro - mandatory workflows, search tool, brainstorming triggers |
| when_to_use | Read this FIRST at start of each conversation when skills are active |
| version | 2.0.0 |
Your personal wiki of proven techniques, patterns, and tools at ~/.claude/skills/.
DO NOT use @ links - they force-load entire files, burning 200k+ context instantly.
INSTEAD, use skill path references:
skills/category/skill-name (no @ prefix, no /SKILL.md suffix)skills/collaboration/brainstorming or skills/testing/test-driven-developmentWhen you see skill references in documentation:
skills/path/name → Use Read tool on ~/.claude/skills/path/name/SKILL.md1. 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:
~/.claude/skills/path/skill-name/SKILL.md"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.
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):
~/.claude/skills/collaboration/remembering-conversations/tool/prompts/search-agent.mdExample:
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.
Every time you start using a skill, announce it:
"I'm using the [Skill Name] skill to [what you're doing]."
Examples:
Why: Transparency helps your human partner understand your process and catch errors early.
If a skill contains a checklist, you MUST create TodoWrite todos for EACH checklist item.
Don't:
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)
Really, try skills-search first.
Categories: skills/INDEX.md → testing, debugging, coding, architecture, collaboration, meta Individual skill: Load from category INDEX
when_to_use match your situation?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.
When writing documentation that references other skills:
Use path format without @ prefix or /SKILL.md suffix:
skills/testing/test-driven-developmentskills/debugging/systematic-debugging@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
Found something valuable? See skills/meta/creating-skills
Want a skill that doesn't exist? Edit skills/REQUESTS.md (at ~/.claude/skills/REQUESTS.md)
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.
In the first response after reading this guide, you MUST announce to the user that you have read the getting started guide
Fork, clone to ~/.clank, run installer, edit CLAUDE.md
RED-GREEN-REFACTOR for process documentation - baseline without skill, write addressing failures, iterate closing loopholes
Interactive idea refinement using Socratic method to develop fully-formed designs
Execute detailed plans in batches with review checkpoints
Execute implementation plan by dispatching fresh subagent for each task, with code review between tasks
Complete feature development with structured options for merge, PR, or cleanup