skill-engineer
Design, test, review, and maintain agent skills for OpenClaw systems using multi-agent iterative refinement. Orchestrates Designer, Reviewer, and Tester subagents for quality-gated skill development. Use when user asks to "design skill", "review skill", "test skill", "audit skills", "refactor skill", or mentions "agent kit quality".
Why use this skill?
Master your agent skill lifecycle with OpenClaw Skill Engineer. Automate design, testing, and maintenance of your agent kit with multi-agent quality-gated workflows.
Install via CLI (Recommended)
clawhub install openclaw/skills/skills/chunhualiao/skill-engineerWhat This Skill Does
The Skill Engineer is the cornerstone of robust OpenClaw agent orchestration. It provides a structured, multi-agent framework to design, build, test, and maintain custom agent skills. By enforcing a strict separation of concerns, it ensures that your agent infrastructure is modular, reliable, and production-ready. The skill acts as an orchestrator, delegating responsibilities to specialized Designer, Reviewer, and Tester subagents to eliminate the 'builder bias' where authors often overlook flaws in their own logic.
It categorizes development into two distinct streams: Capability Uplift Skills (designed to extend model capabilities) and Encoded Preference Skills (designed to enforce organizational workflows). By identifying which category a skill falls into, it applies the appropriate lifecycle management, retirement risks, and testing protocols.
Installation
To install the Skill Engineer, ensure your OpenClaw environment is initialized and run the following command in your terminal:
clawhub install openclaw/skills/skills/chunhualiao/skill-engineer
Before you begin your first design session, you must confirm the availability of mandatory dependencies:
- deepwiki: Essential for querying current OpenClaw API documentation. Check with
ls ~/.openclaw/skills/deepwiki/deepwiki.sh. - Vector memory DB: Enable this in your
openclaw.jsonconfiguration file by settingmemory.enabled: trueto allow semantic searching across your development history and documentation.
Use Cases
- Prototyping New Agent Skills: Rapidly draft, review, and test a new workflow using the multi-agent design cycle.
- Skill Auditing: Perform periodic quality audits on existing skills to determine if they are still necessary (Capability Uplift) or if they require process updates (Encoded Preference).
- Maintenance & Refactoring: Update aging skills that have drifted from organizational requirements or technical API changes.
- Quality Gate Enforcement: Ensure every new skill passes a rigorous Tester subagent examination before being deployed to your primary agent kit.
Example Prompts
- "I need to design a new skill for automated NDA review. Please act as the Skill Engineer and initiate the design phase, classifying it as an Encoded Preference skill."
- "Can you audit my current project skills? I suspect some of the older document-generation skills might be redundant now that our base model has improved."
- "Refactor the compliance-check skill. The legal department has updated the required clauses, so we need to adjust our process fidelity."
Tips & Limitations
- Avoid Self-Review: Always allow the Reviewer and Tester subagents to operate independently. The most common point of failure is when the Designer attempts to skip the review phase.
- Monitor Retirement: If you are building 'Capability Uplift' skills, set a reminder to periodically test against the raw base model. If the model can perform the task natively, it is time to retire the skill to reduce technical debt.
- Deepwiki Reliance: The skill is only as good as your documentation. Ensure your deepwiki repo is kept up-to-date with your latest API changes for the best results.
Metadata
Not sure this is the right skill?
Describe what you want to build — we'll match you to the best skill from 16,000+ options.
Find the right skillPaste this into your clawhub.json to enable this plugin.
{
"plugins": {
"official-chunhualiao-skill-engineer": {
"enabled": true,
"auto_update": true
}
}
}Tags(AI)
Flags: file-read, file-write, code-execution
Related Skills
task-runner
Persistent task queue system. Users add tasks at any time via natural language; tasks are stored in a single persistent queue file and executed asynchronously via subagents. A heartbeat/cron dispatcher wakes periodically to check pending tasks, spawn workers, and report completions. The system never "finishes" — it always remains ready for the next task.
git-repo-to-book
Write a full-length technical book using multi-agent AI orchestration. Spawns parallel research, writing, and review agents to produce 60K-100K+ word manuscripts. Also supports revising individual chapters of existing books. Based on the real workflow that produced an 88,000-word, 14-chapter book in under 18 hours.
claude-usage
Check Claude Max plan usage limits by launching Claude Code and running /usage. Use when the user asks about Claude plan usage, remaining quota, rate limits, or sends /claude_usage.
rose-container-tools
Build and run ROSE compiler tools using ROSE installed in a Docker container. Use when developing source-to-source translators, call graph analyzers, AST processors, or any tool that links against librose.so. Triggers on "ROSE tool", "callgraph", "AST traversal", "source-to-source", "build with ROSE", "librose".
deepwiki
Query DeepWiki MCP to get AI-grounded answers about any public GitHub repository. Use when answering questions about a repo's source code, architecture, configuration, or internals. Triggers on "how does X work in <repo>", "deepwiki", "look up in codebase", "ask deepwiki", "check the source code".