Agent Factory

Standardized bootstrap for 33GOD ecosystem agents. Every agent ships ready to:

• Consume and produce Bloodbank events
• Read/write GOD Docs
• Track work on Plane boards
• Communicate with Cack (boss) and peer agents
• Use hindsight for long-term memory
• Authenticate against all configured providers

Quick Start

When the user requests a new agent, gather these parameters:

Bootstrap Procedure

1. Create Workspace

bash /home/delorenj/.openclaw/skills/33god-agent-factory/scripts/bootstrap.sh \
  --id <id> \
  --name <name> \
  --role <role> \
  --purpose "<purpose>" \
  [--personality "<personality>"] \
  [--model "<model>"] \
  [--workspace-slug "<agent-workspace-slug>"] \
  [--reports-to "<manager>"] \
  [--role-title "<charter title>"] \
  [--directive "<agent-specific directive>"]...

This creates ~/.openclaw/workspace-<id>/ with all template files populated.

2. Role Governance Is Auto-Synced

Bootstrap now calls scripts/role_governance.py automatically to:

• upsert this agent in ~/.openclaw/workspace/frameworks/agent-governance/AGENT_ROLE_MATRIX.json
• apply inherited global prime directives to all managed AGENTS.md files
• apply per-agent directives (when provided via --directive)

You can re-run sync manually at any time:

python3 /home/delorenj/.openclaw/skills/33god-agent-factory/scripts/role_governance.py \
  --governance-dir /home/delorenj/.openclaw/workspace/frameworks/agent-governance \
  apply

3. Update Gateway Config

After running the bootstrap script, update ~/.openclaw/openclaw.json:

Add agent to agents.list:

{
  "id": "<id>",
  "name": "<name>",
  "workspace": "/home/delorenj/.openclaw/workspace-<id>",
  "identity": { "name": "<name>" }
}

If Telegram channel, add to channels.telegram.accounts:

"<account_id>": {
  "name": "<name>",
  "dmPolicy": "pairing",
  "botToken": "<token>",
  "groupPolicy": "allowlist",
  "streamMode": "partial"
}

Add binding to bindings array:

{
  "agentId": "<id>",
  "match": {
    "channel": "telegram",
    "accountId": "<account_id>"
  }
}

4. Restart Gateway

Use the gateway tool with action: "restart" to pick up the new agent.

5. Send Onboarding Briefing

After gateway restart, use sessions_send to the new agent's session (agent:<id>:main) with an onboarding message that includes:

• Their specific mission and current priorities
• Any immediate tasks from the Plane board
• Context about active projects they'll touch

6. Mandatory Process Hardening Check (Required)

Before considering bootstrap complete, verify the new agent understands and acknowledges:

• PR-only delivery (no direct merges to main/master)
• Ticket branch discipline (ticket id in branch name)
• Evidence-first status updates (branch, last commit, git status, stash, open PRs)
• Clean-main handoff after merges (git checkout main && git pull --ff-only)
• BMAD mandatory for coding repos (npx bmad-method@alpha install when missing)

Use sessions_send and require a confirmation response in this structure:

1. Repo hygiene checklist they will run every handoff
2. PR policy acknowledgment
3. BMAD enforcement acknowledgment

Yi Node Flavors (Role Types)

Invariant: Delegator ⇒ persistent memory required.

Base Skills (auto-installed via symlink)

Every agent gets these skills symlinked from Cack's install:

• 33god-creating-and-working-with-projects
• 33god-service-development
• 33god-workflow-generator
• god-docs
• managing-tickets-and-tasks-in-plane
• github
• ecosystem-patterns
• installing-apps-tools-and-services
• hindsight

Template Files

The bootstrap script generates these from templates in references/:

• AGENTS.md — Role-aware instructions + repo execution protocol (ticket branches, PR-only, evidence-first reporting, clean-main handoff, BMAD mandate)
• SOUL.md — Personality + ecosystem identity
• USER.md — Jarad's info (static)
• IDENTITY.md — Agent identity card
• TOOLS.md — Empty, agent fills as needed
• MEMORY.md — Pre-seeded with ecosystem context
• HEARTBEAT.md — Empty (agent configures as needed)
• memory/ directory created

Provider Auth

Providers are configured at the gateway level in agents.defaults, so all agents automatically inherit:

• anthropic, github-copilot, openai-codex, opencode, kimi-coding, google, google-antigravity, openrouter

No per-agent provider config needed — it's all in defaults.

Conventions

• Workspace: ~/.openclaw/workspace-<id>/
• Session key: agent:<id>:main
• Telegram account: matches agent id unless overridden
• All agents know Cack is the coordinator/boss
• All agents use sessions_send for inter-agent comms
• Plane workspace: 33god

Per-Agent Directive Updates (Post-Bootstrap)

Use this when you want to modify one agent's local directives later:

python3 /home/delorenj/.openclaw/skills/33god-agent-factory/scripts/role_governance.py \
  --governance-dir /home/delorenj/.openclaw/workspace/frameworks/agent-governance \
  upsert-and-apply \
  --workspace /home/delorenj/.openclaw/workspace-<id> \
  --agent "<name>" \
  --role "<role title>" \
  --reports-to "<manager>" \
  --mission "<mission>" \
  --directive "<local directive 1>" \
  --directive "<local directive 2>"