Agent skill

project-init

Use when setting up a new project for AI-agent-driven management, connecting GitHub as a command center, or when user says "init project", "настрой проект", "set up my project", "initialize". First skill in the Personal Corp framework.

Stars 136
Forks 14

Install this agent skill to your Project

npx add-skill https://github.com/serejaris/ris-claude-code/tree/main/skills/project-init

SKILL.md

Project Init

Set up GitHub as your business operating system through a guided interview. Part of the Personal Corp framework — a system for running a business as one person with AI agents.

What This Does

Interviews you about your work, then creates the infrastructure for agent-driven project management:

  • GitHub Project with the right fields
  • Labels for retro/planning cycles
  • CLAUDE.md config block so other skills (weekly-retro, weekly-planning) know your setup
  • Canonical files for single-source-of-truth data

Phase 1: Interview (one question at a time)

Question flow

dot
digraph init {
    "What do you do?" -> "What repos do you use?";
    "What repos do you use?" -> "What are your products/services?";
    "What are your products/services?" -> "Where do you track numbers?";
    "Where do you track numbers?" -> "What's your weekly rhythm?";
    "What's your weekly rhythm?" -> "Generate config";
}
  1. What do you do? — Business type, role, solo or team. Keep it conversational.
  2. What repos do you use? — List them. Check with gh repo list --limit 30. Categorize: execution, growth, infrastructure.
  3. What are your products/services? — What you sell, pricing, where it lives.
  4. Where do you track numbers? — Revenue, users, metrics. This becomes your canonical files list.
  5. What's your weekly rhythm? — Delivery days, meetings, deep work blocks. This shapes retro/planning cadence.

Rules

  • One question at a time — don't overwhelm
  • Verify what existsgh project list, gh label list, check for existing CLAUDE.md
  • Don't overwrite — if CLAUDE.md exists, ADD config block, don't replace

Phase 2: Generate Config

Create the config block for CLAUDE.md:

markdown
## Agent Operations Config

### Repos (scanned during retro/planning)
repos:
  - owner/repo-1    # execution — what it does
  - owner/repo-2    # growth — what it does
  - owner/repo-3    # infrastructure — what it does

### GitHub Project
project_id: N       # gh project list --owner {owner}
owner: {github-username}

### Canonical Files (single source of truth)
canonical:
  numbers: data.md          # revenue, users, metrics
  products: product.md      # current offers and pricing
  insights: insights.md     # lessons learned

### Task Routing (which issues go where)
routing:
  - pattern: "bot, broadcast, onboarding"
    repo: owner/bot-repo
  - pattern: "content, lessons"
    repo: owner/content-repo
  - pattern: "strategy, cross-cutting"
    repo: owner/main-repo

### Retro/Planning
retro_label_prefix: "retro"    # creates retro:W{NN}
retro_log_path: "docs/activity-log/"
weekly_cadence: "Thursday"      # main delivery day

### Interview Topics (for weekly-retro)
interview_order:
  - "Delivery — main product/service output"
  - "Clients/Users — new, churned, notable"
  - "Calendar events — verify each one"
  - "New initiatives — what started, why"
  - "Open question — anything else?"

Show the generated config to the user. Ask: "Does this look right?"

Phase 3: Create Infrastructure

After user confirms:

bash
# 1. Create GitHub Project (if doesn't exist)
gh project create --owner {owner} --title "{Project Name}"

# 2. Add custom fields
# Status: Todo / In Progress / Done
# Deadline: date
# Week: single-select (configurable)

# 3. Create labels in each repo
for repo in {repos}; do
  gh label create "retro:W00" -R $repo --color "D4C5F9" --description "Weekly retro backlog template"
  gh label create "epic" -R $repo --color "FF6B6B" --description "Epic / initiative"
done

# 4. Create canonical files (if they don't exist)
# Only create stubs — don't fill with fake data

# 5. Append config block to CLAUDE.md

Phase 4: Verify

bash
# Check project exists
gh project list --owner {owner}

# Check labels
for repo in {repos}; do
  gh label list -R $repo | grep retro
done

# Check CLAUDE.md has config
grep "Agent Operations Config" CLAUDE.md

Show results. Confirm everything works.

What Happens Next

After init, you can use:

  • weekly-retro — structured retrospective using your config
  • weekly-planning — prioritize backlog into weekly outcomes

Each skill reads the config block from CLAUDE.md — no re-configuration needed.

Red Flags — STOP

  • Existing CLAUDE.md with content → ADD config block, don't overwrite
  • Existing GitHub Project → ask if they want to use it or create new
  • User unsure about repos → gh repo list and walk through together
  • User doesn't use GitHub → this framework requires GitHub. Suggest alternatives or help set up.

Common Questions

"I have 20+ repos" — Pick the 5-7 that matter for weekly operations. The rest can be added later.

"I don't track numbers anywhere" — Start with one file. Even data.md with three lines is better than nothing. The retro skill will remind you to update it.

"My project structure is messy" — That's fine. Init creates a clean layer on top. You don't need to reorganize first.

Didn't find tool you were looking for?

Be as detailed as possible for better results