npx bkit-doctor setupThat's it. One command scans your project, fixes what's broken, and wires Claude Code to document everything automatically — forever.
See what setup does under the hood
bkit-doctor setup
[1/4] 🔍 Smart Check — scanning for missing AI configs...
✔ .claude/ directory found
✔ CLAUDE.md found
⚠ hooks.json missing → will fix
[2/4] 🏗️ Interactive Init — scaffolding missing structure...
✔ hooks.json created
✔ settings.local.json created
✔ docs/ scaffolded
[3/4] 🛠️ Auto-Fix — applying CLAUDE.md...
✔ CLAUDE.md written (backup: CLAUDE_20260330_backup.md)
[4/4] 🤖 Skill Injection — generating SKILL.md + npm scripts...
✔ SKILL.md created
✔ Added to package.json: bkit:check, bkit:fix, bkit:setup
Setup complete. Claude Code will now follow PDCA workflows automatically.
After setup, use the npm shortcuts:
npm run bkit:check # diagnose your project
npm run bkit:fix # auto-fix all issues
npm run bkit:setup # re-run the wizard anytimeIdempotent & CI-safe. Running
setuptwice is always safe. In non-TTY environments (CI/CD), interactive prompts are skipped and existing files are preserved.
setup injects a SKILL.md into your project root. Claude Code reads it as project context and follows three rules on every task, automatically.
graph LR
A[💬 Your Prompt] --> B[Claude reads SKILL.md]
B --> C{Rule match?}
C -->|RULE 1| D[📋 pdca-plan auto-runs]
C -->|RULE 2| E[🔍 pdca-list state check]
D --> F[⚙️ You code with Claude]
E --> F
F -->|RULE 3| G[📝 pdca-do/check/report auto-runs]
G --> H[✅ Audit trail built]
The three rules injected into your project:
| Rule | Trigger | Action |
|---|---|---|
| RULE 1: PROACTIVE DOCUMENTATION | Before writing code | Auto-runs pdca-plan to create a structured plan |
| RULE 2: STATE SYNC | Before writing code | Checks existing PDCA state with pdca-list |
| RULE 3: PIPELINE | After coding | Auto-runs pdca-do → pdca-check → pdca-report |
Claude Code reads
SKILL.mdas project context. No plugin installation required — it works out of the box.
Result: Every feature, bugfix, and refactor is automatically planned, executed, verified, and reported — building a permanent audit trail with zero manual overhead.
For power users who want individual control over each step that setup orchestrates.
npx bkit-doctor check [--path <dir>]Runs 16 diagnostic checks across your project structure and reports pass, warn, or fail for each. Exits with code 1 on hard failures — CI-friendly.
| Category | Checks | Severity |
|---|---|---|
| structure | .claude/ directory |
hard (exit 1) |
| config | CLAUDE.md |
hard (exit 1) |
| config | hooks.json, settings.local.json |
soft |
| agents | 4 agent definition files | soft |
| skills | 7 skill files under .claude/skills/ |
soft |
| policies & templates | 4 + 4 files | soft |
| docs | docs/01-plan/ → docs/04-report/, output/pdca/ |
soft |
| changelog | CHANGELOG.md |
soft |
npx bkit-doctor init [--preset <name>] [--target <name>] [--yes]Scaffolds specific targets or full presets.
| Target | Creates |
|---|---|
claude-root |
.claude/ directory |
hooks-json |
.claude/hooks.json |
settings-local |
.claude/settings.local.json |
agents-core |
4 agent definition files |
skills-core |
7 SKILL.md files under .claude/skills/ |
templates-core |
4 document templates |
policies-core |
4 policy files |
docs-core |
All docs/ directories |
docs-pdca |
output/pdca/ directory |
docs-changelog |
CHANGELOG.md |
Presets: default (full) · lean (minimal) · workflow-core · docs
npx bkit-doctor fix [--yes] [--dry-run]Runs check → recommend → init in sequence. Use --dry-run to preview what would change before committing. Use --yes to skip confirmation prompts.
npx bkit-doctor skill [--append-claude] [--overwrite] [--stdout] [--dry-run]Generates SKILL.md with the three PDCA automation rules. Use --append-claude to inject the rules directly into CLAUDE.md instead.
npx bkit-doctor clear [--path <dir>]
⚠️ Confirmation required. This command interactively lists bkit-doctor generated files and asks for explicit confirmation before deleting anything. No silent data loss.
Generate structured Plan-Do-Check-Act documents for any task. State is tracked in .bkit-doctor/pdca-state.json.
# Generate a full guide in one shot
npx bkit-doctor pdca "User Authentication" --type feature --owner alice --priority P1
# Or stage-by-stage
npx bkit-doctor pdca-plan "User Authentication"
npx bkit-doctor pdca-do "User Authentication"
npx bkit-doctor pdca-check "User Authentication"
npx bkit-doctor pdca-report "User Authentication"
# List all tracked topics
npx bkit-doctor pdca-listDocument types: guideline · feature · bugfix · refactor
Output: output/pdca/<slug>-pdca-{stage}.md — versioned, auditable, git-trackable.
bkit-doctor enforces structure and injects automation rules. bkit is the Claude Code plugin that runs the AI workflow engine inside Claude Code.
| bkit-doctor | bkit (plugin) | |
|---|---|---|
| Project structure | ✅ creates & validates | — |
| CLAUDE.md / SKILL.md | ✅ generates | reads |
| PDCA document engine | ✅ file generation | orchestration |
| AI agents & skills | — | ✅ 31 agents / 36 skills |
| Runs in | terminal | Claude Code |
# Install bkit inside Claude Code
/plugin marketplace add popup-studio-ai/bkit-claude-code# GitHub Actions
- name: Check project structure
run: npx bkit-doctor check
# Exits 1 if .claude/ or CLAUDE.md is missingContributions are welcome. Please open an issue first to discuss what you'd like to change.
- Fork the repository
- Create a feature branch:
git checkout -b feat/my-feature - Run tests:
npm test - Submit a pull request
Apache-2.0 © dotoricode