openclaw-dev

Guided workflow to create a new OpenClaw skill from scratch. Walks through requirements, design, implementation, validation, and deployment.

Deploy an OpenClaw skill to an agent's workspace, restart gateway if needed, trigger new session, and verify the skill loaded.

Systematic OpenClaw runtime diagnosis — analyze logs, identify fault patterns, output report, and accumulate new findings. Supports remote Gateways.

Analyze OpenClaw session logs and evolve a skill based on usage data

Validate openclaw.json syntax and semantics before applying — prevent configuration corruption

You are a capability evolution agent for the `openclaw-dev` toolkit. Your job is to diagnose and improve OpenClaw capabilities using the bundled QA module at `plugins/qa`, then verify regressions.

Use this agent when an OpenClaw plugin or compatible bundle has been created or modified and needs comprehensive validation. Triggers: 'validate my plugin', 'check plugin structure', 'review openclaw.plugin.json', 'plugin quality check'.

OpenClaw skill reviewer agent. Use this agent to review an OpenClaw SKILL.md for quality, correctness, and best practices compliance. Checks frontmatter, description effectiveness, body structure, progressive disclosure, and context efficiency.

Use this skill when the user asks to design, implement, audit, or evolve model routing policies across multiple providers/models. Triggers: 'Which model should I use for X?', 'Design a routing policy across many models/providers', 'Work/private/sensitive routing policy', 'Coding should prefer domestic models, but escalate on hard tasks', 'How do we handle future model name changes?', 'Create a router script / policy config / fallback chain', 'Audit current model routing', work/private routing, coding vs research routing, provider safety constraints, fallback chains, slot/alias-based routing, or future-proof model name migration. Includes a runnable router and policy validator.

Use this skill when the user asks about OpenClaw architecture overview, how OpenClaw works internally, session model, channel concepts, workspace structure, agent routing internals, plugin API design (openclaw.plugin.json, api.register*), agent configuration schema (agents.list[], bindings), multi-agent delegation model, SOUL.md/AGENTS.md/USER.md persona design, memory search internals, cron/heartbeat mechanisms, hooks/webhooks architecture, browser automation model, tool policy design, sandbox security model, node pairing protocol, discovery protocol, CLI command reference, source code structure, building from source, testing, releasing, or any OpenClaw internals/theory question. Also use for: 'create an agent', 'scaffold agent', 'create a plugin', 'scaffold plugin', 'sync knowledge', 'update knowledge base'. For hands-on operations (install, debug, configure, fix, diagnose, set up networking) use openclaw-node-operations instead.

Use this skill when asked to install OpenClaw, set up a node, configure a Gateway, onboard a new machine, debug OpenClaw issues (read logs, run doctor, health checks, diagnose faults), fix Gateway problems, set up networking (Tailscale, SSH tunnels), check node status, troubleshoot connectivity, configure remote access, deploy on Linux/Windows/macOS, lint config, validate openclaw.json, check fleet status, query agent/channel/plugin status, or run systematic diagnostics. Also use for: 'diagnose OpenClaw', 'lint my config', 'validate configuration', 'show status', 'fleet status', 'Gateway health', 'check OpenClaw health'. Covers hands-on operations: installation, onboarding, Gateway service management, remote access, cross-OS support, debugging, monitoring, diagnostics, config validation. For architecture/theory questions use openclaw-dev-knowledgebase instead.

Guided workflow for creating, validating, deploying, and evolving OpenClaw skills. Use when asked to 'create a skill', 'build a SKILL.md', 'develop an agent skill', 'validate skill', 'deploy skill', 'evolve skill', 'improve skill from usage data', 'analyze session logs for skill performance', 'upgrade skills based on usage', 'list skills', 'show installed skills', 'what skills are available', or any skill lifecycle task. Covers: requirements, scaffolding, frontmatter, deployment, session verification, data-driven evolution, and skill inventory.

Internal security agent that intercepts and evaluates sub-agent command approval requests. Analyzes commands for safety, auto-approves safe operations, and provides remediation guidance for dangerous commands. This is a middleware agent - it doesn't interact with users directly, only with requesting sub-agents.

1 hook across 1 event