openclaw-ruh

Playbook for designing and shipping high-quality digital employees on the openclaw-ruh-enterprise platform. Use when helping the user build a new agent, improve the Architect agent itself, review an architecture.json or SKILL.md, or refactor an agent's workspace files (SOUL.md, PRD.md, TRD.md, skills/*/SKILL.md). Covers the full Think → Plan → Build → Review → Test → Ship → Reflect pipeline and encodes the Claude Code patterns (progressive disclosure, description-first matching, strict tool scopes) that produce agents that actually work in production.

Playbook for designing and shipping high-quality digital employees on the openclaw-ruh-enterprise platform. Use when helping the user build a new agent, improve the Architect agent itself, review an architecture.json or SKILL.md, or refactor an agent's workspace files (SOUL.md, PRD.md, TRD.md, skills/*/SKILL.md). Covers the full Think → Plan → Build → Review → Test → Ship → Reflect pipeline and encodes the Claude Code patterns (progressive disclosure, description-first matching, strict tool scopes) that produce agents that actually work in production.

Manage the production GCP infrastructure for openclaw-ruh-enterprise. SSH into VMs, check resource usage (CPU, RAM, disk), view Docker containers, read logs, restart services, clean up disk, and run server maintenance. Use when asked to "check the server", "server health", "disk usage", "restart backend", "clean docker", "server logs", "SSH into prod", or any production infrastructure task.

Read logs and telemetry for openclaw-ruh-enterprise in local dev. Backend HTTP + gateway-proxy logs (/tmp/backend.log), per-sandbox gateway logs (/tmp/openclaw-gateway.log inside each container), OpenClaw structured JSON subsystem logs (/tmp/openclaw/openclaw-YYYY-MM-DD.log), Postgres state queries, and richer telemetry (Langfuse, OTEL, GlitchTip) when available. Use when asked to "check the logs", "why is the agent stuck", "what's the backend doing", "grep the gateway log", "look at the trace", or any local-dev log investigation task. For production (GCP VM) logs, use the /gcp-server skill instead.

Knowledge base maintenance and agent access for openclaw-ruh-enterprise. Five modes: read (orient before any task), spec (create a feature spec), link (verify and fix wikilink graph), audit (full KB health check), update (update existing notes after code changes). Use when asked to "check the KB", "create a spec", "update the knowledge base", "audit KB health", or at the start of any development task. Proactively suggest at Think phase and Ship phase of the gstack sprint.

Create and monitor Hermes tasks through the queue API with reproducible curl commands.

Knowledge base maintenance and agent access for openclaw-ruh-enterprise. Five modes: read (orient before any task), spec (create a feature spec), link (verify and fix wikilink graph), audit (full KB health check), update (update existing notes after code changes). Use when asked to "check the KB", "create a spec", "update the knowledge base", "audit KB health", or at the start of any development task. Proactively suggest at Think phase and Ship phase of the gstack sprint.

Analyze developer automation requirements and decompose them into a skill graph, Lobster workflow, and OpenClaw agent config.

Search ClawHub and skills.sh for existing OpenClaw skills by keyword.

Interact with the aget-data-ingestion FastAPI service: ingest data from external sources, query entity/result tables, write computed results. MANDATORY for every generated agent system.

Provision isolated Daytona sandboxes for generated multi-agent systems. Create sandboxes, install OpenClaw, deploy workspace files, configure environment variables, and start the gateway.

File system operations for the builder agent: create directories, write files, copy templates, read file contents. Used when generating multi-agent system output directories.

Git operations for initializing repos, committing files, and managing branches. Used by builder and deployer agents when preparing generated systems for GitHub push.

GitHub API operations via gh CLI: create repositories, manage collaborators, create releases. Used by the deployer agent to publish generated systems.

Generate Lobster workflow YAML files from a skill graph DAG. Creates deterministic, step-by-step pipelines with proper dependencies, approval gates, and mandatory write_results terminal step.

Validate generated multi-agent systems: check SKILL.md syntax, verify openclaw.json config, run Lobster workflows against test data, verify result writes to data-ingestion service.

Interact with the aget-data-ingestion FastAPI service: ingest data from external sources, query entity/result tables, write computed results. MANDATORY for every agent.