ClaudeKit usage guide — discover commands, skills, workflows naturally.
All-in-one ClaudeKit guide. Run the script and present output based on type markers.
The script uses keyword matching with smart weighting. After getting results, validate against these heuristics:
| Sentence Pattern | Primary Intent | Example |
|---|---|---|
[action verb] my [object] | The action verb | "commit my changes" → git |
[context] [subject noun] | The subject noun | "setup notifications" → notifications |
[noun] [noun] | Last noun (topic) | "discord webhook" → notifications |
Action verbs (high intent when first): fix, test, commit, push, build, create, review, deploy, run, check, find, plan, refactor
Context words (low intent, modify subject): setup, add, start, new, my, the, configure
Override script only if: result clearly mismatches the sentence pattern above. Otherwise trust the algorithm.
IMPORTANT: Always translate $ARGUMENTS to English before passing to script.
The Python script only understands English keywords. If $ARGUMENTS is in another language:
$ARGUMENTS to Englishpython .claude/skills/ck-help/scripts/ck-help.py "$ARGUMENTS"
The script outputs a type marker on the first line: @CK_OUTPUT_TYPE:<type>
Read this marker and adjust your presentation accordingly:
@CK_OUTPUT_TYPE:comprehensive-docsFull documentation (config, schema, setup guides).
Presentation:
@CK_OUTPUT_TYPE:category-guideWorkflow guides for skill categories.
Presentation:
@CK_OUTPUT_TYPE:command-detailsSingle skill/subcommand documentation.
Presentation:
@CK_OUTPUT_TYPE:search-resultsSearch matches for a keyword.
Presentation:
@CK_OUTPUT_TYPE:task-recommendationsTask-based skill suggestions.
Presentation:
Script output = foundation. Your additions = value-add.
Never replace or summarize the script output. Always show it fully, then enhance with your knowledge and context.
/ck:plan → /ck:cook: Best for high-risk or complex changes/ck:cook: Standalone for straightforward implementation/ck:plan is mandatory before /ck:cook