A specialized tool for generating automated test suites and SKILL.md files to evaluate and improve AI agent capabilities.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install auto-skill-forge
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install auto-skill-forge using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Skill Forge?
Skill Forge is a powerful developer utility designed to streamline the lifecycle of creating and validating Openclaw Skills. By performing static analysis on existing documentation or processing structured specifications, it automatically produces a task_suite.yaml that functions as a robust test harness. This allows developers to move beyond manual scaffolding and enter a data-driven improvement loop.
The tool is particularly effective at turning theoretical capabilities into measurable benchmarks. Whether you are adding tests to an existing skill or forging a new one from scratch, Skill Forge ensures your agent's performance is quantifiable. It bridges the gap between writing a skill and knowing it actually works by extracting scenarios from frontmatter, use cases, and code examples to build high-quality Openclaw Skills.
Skill Forge Use Cases
- Adding automated test suites to an existing Skill that lacks a task_suite.yaml.
- Generating a complete SKILL.md and test harness from a minimal skill_spec.yaml description.
- Executing a full generate-evaluate-improve lifecycle in a single command pass.
- Validating the test coverage of manually written suites against automatically generated scenarios.
How Skill Forge Works
- The tool reads a SKILL.md file or a structured specification as input.
- It performs static analysis to extract testable claims from five distinct sources: frontmatter, use cases, examples, anti-examples, and output formats.
- For each extracted scenario, the engine assigns an appropriate judge type, such as LLMRubricJudge, ContainsJudge, or PytestJudge.
- It performs null-skill calibration to ensure the tests specifically measure the unique value added by Openclaw Skills rather than baseline model knowledge.
- The final tasks are deduplicated, filtered for quality, and exported as a human-readable YAML file.
Skill Forge Setup
To use Skill Forge, you must have the script environment configured. You can generate suites for existing Openclaw Skills using the following commands:
# Generate a test suite for an existing skill
python3 scripts/forge.py --from-skill /path/to/skill-dir --output /path/to/output
# Generate and immediately evaluate performance
python3 scripts/forge.py --from-skill /path/to/skill-dir --output /path/to/output --evaluate
Skill Forge Data Schema & Taxonomy
The tool generates structured artifacts to support the development of Openclaw Skills:
| Artifact | Purpose | Format |
|---|---|---|
| task_suite.yaml | A collection of 5-10 test tasks with prompts and judges | YAML |
| SKILL.md | The full skill documentation generated from a spec | Markdown |
| evaluation_report.json | Detailed pass/fail metrics and aggregate scores | JSON |
| Improved SKILL.md | The updated skill document after an auto-improve cycle | Markdown |
Skill Forge Advanced Features
- Null-skill calibration that filters out trivial tasks to ensure meaningful performance metrics.
- Execution-harness pattern detection for scripted tools, checking for atomic writes and error handling.
- Full lifecycle integration with improvement-orchestrator for automated multi-round self-improvement.
- Support for mock LLM modes to perform heuristic calibration without API costs.
SKILL.md
Loading
Related Openclaw Skills
Skill Creator
A professional toolkit for authoring, auditing, and packaging high-quality agent skills with a focus on executable precision.
UI Automation Test Agent
A sophisticated agent skill that captures browser actions to generate replayable test scripts and visual HTML reports automatically.
Deprecated Skill
A legacy component within the ecosystem that has been retired and is no longer supported for active development workflows.
Iceberg Excavation
A deep-dive psychological analysis tool that transforms emotional triggers into profound self-awareness and systemic behavioral change.
KeyAPI YouTube Channel Analysis
A robust intelligence toolset for profiling YouTube channels, extracting deep metadata, and performing advanced filtered searches.
TikTok Shop Creator Discovery
Identify top-performing TikTok Shop creators and analyze their sales performance, audience demographics, and GMV metrics.
Featured*
