onboard

name	onboard
description	Interactive project setup: copy templates, configure fleet, define compliance floor, add first specialist agent.

Onboard

Interactive setup for a new project using the Venutian Antfarm harness. Follows docs/GETTING-STARTED.md.

Workflow

Step 1: Check Prerequisites

Verify required tools are available:

command -v git && command -v claude && command -v bash
command -v jq  # optional but recommended

Report any missing prerequisites before proceeding.

Step 1b: Scaffold Runtime Directories

Create the runtime directories and copy content templates to their runtime locations. These directories are not shipped in the repo — they are populated by /onboard at setup time.

# Runtime directories (empty, populated by agents at runtime)
mkdir -p .claude/compliance/{compiled,eslint,semgrep,proposals}
mkdir -p .claude/governance/decisions
mkdir -p .claude/governance/guidance/{ceo,cfo,ciso,cko,coo,cto}
mkdir -p .claude/floors
mkdir -p .claude/findings
mkdir -p .claude/metrics

# Copy content templates to runtime locations
cp templates/compliance/change-log.md .claude/compliance/change-log.md
cp templates/compliance/targets.md .claude/compliance/targets.md
cp templates/governance/executive-brief.md .claude/governance/executive-brief.md
cp templates/governance/guidance-registry.md .claude/governance/guidance-registry.md
cp templates/findings/register.md .claude/findings/register.md
cp templates/findings/information-needs.md .claude/findings/information-needs.md

# Initialize empty runtime files
touch .claude/metrics/events.jsonl

Note: The CRO does not have a guidance directory — CRO uses the compliance infrastructure (proposals, change-log) rather than a guidance slot.

Step 2: Compliance Floor

Check if floors/compliance.md exists at project root.

• If missing: Copy from templates/floors/compliance.md and ask the user to define 3-5 non-negotiable rules for their domain. Provide examples:
  • E-commerce: "No credit card numbers stored outside the payment processor"
  • Healthcare: "No PHI transmitted without encryption"
  • SaaS: "Tenant data isolation enforced on every query"
• If exists: Read it, confirm with user, move on.

Step 2b: Governance Activation

After the compliance floor is defined:

1. CRO takes guardianship. Generate initial checksum:

   mkdir -p .claude/floors/compliance
   sha256sum floors/compliance.md | cut -d' ' -f1 > .claude/floors/compliance/floor-checksum.sha256
   echo "$(git rev-parse HEAD)" >> .claude/floors/compliance/floor-checksum.sha256
   

2. CISO security review. Dispatch the CISO agent to evaluate whether the floor adequately covers security for the project's domain. The CISO may propose additions via /compliance propose.
3. Process proposals. If the CISO proposed additions, the CRO processes them through the standard change control process. User approves the final floor.

Step 2c: Behavioral Floor Setup

Ask the user if they want to define behavioral rules for their team.

• If yes: Copy from templates/floors/behavioral.md to floors/behavioral.md. Walk the user through example behavioral rules:
  • "We MUST ALWAYS run full validation before handoff"
  • "We MUST NEVER skip the findings loop on notable events"
  • "We MUST ALWAYS confirm intent before designing for non-trivial tasks"
• Help the user define 2-4 behavioral rules appropriate for their team's workflow.
• Generate initial checksum:

  mkdir -p .claude/floors/behavioral
  sha256sum floors/behavioral.md | cut -d' ' -f1 > .claude/floors/behavioral/floor-checksum.sha256
  echo "$(git rev-parse HEAD)" >> .claude/floors/behavioral/floor-checksum.sha256
  

• If no: Skip. The behavioral floor can be added later via /behavioral propose.

Step 3: Fleet Configuration

Check if fleet-config.json exists at project root.

• If missing: Copy from templates/fleet-config.json and walk the user through:
  • project.name — their project name
  • metrics.backend — recommend "jsonl" to start
  • agents.specialists — which specialist agents they plan to add
  • retro.cadence — recommend 1 (retro after every item) for new teams
• If exists: Read it, confirm with user, move on.

Step 4: First Specialist Agent

Ask the user what their primary tech stack is, then recommend a specialist template:

Stack	Template
Node/Python/Go/Rust backend	templates/agents/backend-specialist.md
React/Vue/Angular frontend	templates/agents/frontend-specialist.md
Security-focused	templates/agents/security-reviewer.md
E2E testing	templates/agents/e2e-test-engineer.md
Infrastructure/DevOps	templates/agents/infrastructure-ops.md

Copy the chosen template to .claude/agents/ and help the user customize:

• Domain description
• Responsibilities
• Tech stack specifics

Step 5: First Backlog Item

Help the user create their first tier file at docs/plans/tier-1-launch-blockers.md with at least one work item. Walk them through the item format defined in .claude/COLLABORATION.md.

Step 6: Validate

Run validation to confirm everything is wired up:

ops/dora.sh          # metrics pipeline works
ops/pathways.sh      # pathway analysis works
bash -n ops/*.sh     # all scripts parse

Step 7: Summary

Print a checklist of what was set up:

## Onboarding Complete

- [x] Compliance floor defined
- [x] Fleet configured
- [x] Specialist agent added: <name>
- [x] First backlog item created
- [x] Validation passed
- [x] Compliance floor guardianship activated (CRO + checksum)
- [x] Security review completed (CISO)
- [ ] Behavioral floor defined (optional)

Next steps:
- Run `/po` to see your status overview
- Run `/po groom` to refine your first item
- Start building with `/po next`