tasuki-onboard

name	tasuki-onboard
description	Scan a codebase, detect the stack, and auto-generate a complete .tasuki/ configuration with agents, rules, hooks, skills, and TASUKI.md.
allowed-tools	Read, Write, Edit, Glob, Grep, Bash

Tasuki Onboard — Project Discovery & Configuration

Run the full onboarding pipeline for the current project.

Step 1: SCAN — Detect the stack

Run each detection script and capture the JSON output:

bash tasuki/src/detectors/detect-backend.sh .
bash tasuki/src/detectors/detect-frontend.sh .
bash tasuki/src/detectors/detect-database.sh .
bash tasuki/src/detectors/detect-infra.sh .
bash tasuki/src/detectors/detect-testing.sh .

Collect the results into a detection summary.

Step 2: ANALYZE — Match against profiles

1. Read tasuki/src/profiles/ to find the best-fit stack profile
2. Read tasuki/src/registry.yaml to determine which agents to activate
3. Decide which agents to skip based on stack_required fields:
   • No frontend detected? Skip frontend-dev
   • No database detected? Skip db-architect
   • No Docker detected? Simplify devops
   • Always include: planner, qa, security, reviewer

Step 3: ANALYZE — Study the codebase

Read ~10-15 key files to understand project-specific conventions:

• Entry point (main.py, app.ts, etc.)
• 2-3 existing route/controller files (to learn patterns)
• 1-2 model/schema files
• Auth middleware or dependency
• Existing test files (to learn test patterns)
• Docker/compose files
• CI/CD config

Extract:

• Project name
• Project description
• File path conventions
• Auth pattern
• Coding conventions not in the profile

Step 4: GENERATE — Write configuration

Using the templates from tasuki/src/templates/ and the profile conventions:

1. .tasuki/agents/*.md — One file per activated agent, populated with project context
2. .tasuki/rules/*.md — One file per detected file type, with project conventions
3. .tasuki/hooks/protect-files.sh — Copy from template, optionally customize
4. .tasuki/hooks/security-check.sh — Copy from template
5. .tasuki/settings.json — Generate permissions for detected paths + tools
6. .mcp.json — Suggest MCP servers based on profile
7. TASUKI.md — The complete orchestration brain, using the TASUKI.md template

Replace all {{PLACEHOLDER}} values in templates with actual project values.

Step 5: VERIFY

1. Read back all generated files
2. Check that agent names in TASUKI.md match agent files
3. Check that rule paths match actual project structure
4. Present a summary:

Tasuki Onboarding Complete
==========================

Project: {name}
Stack: {backend} + {frontend} + {database}

Agents activated (N/9):
  [x] planner
  [x] qa
  [x] db-architect
  [x] backend-dev
  [ ] frontend-dev (skipped: no frontend detected)
  [x] security
  [x] reviewer
  [x] devops

Rules: {list}
Hooks: protect-files, security-check
Skills: {list}
MCP Servers: {list}

Ready to test? Try: "Plan a small feature to verify the pipeline"