skill-builder
Scaffold a new ClawBio skill from a spec file (JSON/YAML) or interactively — generates SKILL.md, Python skeleton, tests, and updates catalog.json
- License 许可
- MIT
- 最近更新
- 2026-11-14
- Stars
- ★ 12
- Forks
- 5
01 前言 Front Matter
- name
- skill-builder
- license
- MIT
- metadata
- { "tags": [ "meta", "scaffolding", "developer-tools", "skill-creation" ], "author": "Mj", "version": "0.1.0", "openclaw": { "os": [ "darwin", "linux" ], "emoji": "🦖", "always": false, "install": null, "homepage": "https://github.com/ClawBio/ClawBio", "requires": { "env": null, "bins": [ "python3" ], "config": null }, "trigger_keywords": [ "create skill", "new skill", "scaffold skill", "skill template", "skill builder", "add a skill", "build a skill", "make a skill" ] } }
- description
- Scaffold a new ClawBio skill from a spec file (JSON/YAML) or interactively — generates SKILL.md, Python skeleton, tests, and updates catalog.json
02 资源文件 4
- 参考文档demo_spec.json1.2 KB
- 脚本skill_builder.py72.3 KB
- 参考文档templates/spec_template.json3.0 KB
- 脚本tests/test_skill_builder.py44.2 KB
03 SKILL.md 正文
🦖 Skill Builder
[!note] Vault audit 2026-07-24 — USE-10 Use this to scaffold a brand-new skill from a spec or interactively; to auto-draft skills from observed screen workflows use
autoskill, to eval-tune an existing skill against an LLM-judge useclawpathy-autoresearch, to package a whole plugin bundle useplugin-creator. Distinguishing axis: authoring mode (manual scaffold vs observation vs eval-tuning vs plugin packaging).
You are Skill Builder, a specialised ClawBio meta-skill for scaffolding new skills. Your role is to take a skill specification and generate a complete, PR-ready ClawBio skill directory with all required files.
Why This Exists
- Without it: Contributors must manually copy the template, fill in every section, write a Python skeleton from scratch, and manually update
catalog.jsonandclawbio.py— a 30-60 minute process prone to missing required sections or malformed YAML. - With it: Provide a JSON spec and get a complete, validated, immediately runnable skill scaffold in seconds, ready to submit as a pull request.
- Why ClawBio: The scaffold enforces all requirements from
CONTRIBUTING.mdautomatically — no forgotten sections, no malformed frontmatter, no missing reproducibility bundle.
Core Capabilities
- Spec-driven scaffolding: Read a JSON (or YAML with pyyaml) spec file and generate a complete skill directory.
- Interactive mode: Prompt for skill details when no spec file is provided (
--interactive). - Validation: Check any existing
SKILL.mdagainst the CONTRIBUTING.md checklist (--validate-only). - Auto-registration: Update
skills/catalog.jsonand patchclawbio.py'sSKILLSdict when run from inside the ClawBio repo. - Dry-run preview: Print all generated content without writing files (
--dry-run).
Input Formats
| Format | Extension | Required Fields | Example |
|---|---|---|---|
| JSON spec | .json |
name, description, author | spec.json |
| YAML spec | .yaml / .yml |
name, description, author | spec.yaml (requires pyyaml) |
| Existing SKILL.md | .md |
Any SKILL.md | Used with --validate-only |
Workflow
When the user asks to create a new skill:
- Load spec: Read JSON/YAML spec file, or collect fields interactively if
--interactive - Validate spec: Check required fields (name, description, author); apply defaults for optional fields
- Generate files: Create
SKILL.md,<name>.py,tests/test_<name>.py,examples/example_spec.json - Update registry: If repo root found, append entry to
catalog.jsonand patchSKILLSdict inclawbio.py - Report: Print a summary of generated files and next steps
CLI Reference
# Spec-driven (recommended for agents)
python skills/skill-builder/skill_builder.py --input spec.json --output skills/my-skill/
# Interactive (human-friendly)
python skills/skill-builder/skill_builder.py --interactive
# Demo (scaffolds hello-bioinformatics skill)
python skills/skill-builder/skill_builder.py --demo --output /tmp/skill_builder_demo
# Validate an existing SKILL.md
python skills/skill-builder/skill_builder.py --validate-only --input skills/my-skill/SKILL.md
# Dry run (print without writing)
python skills/skill-builder/skill_builder.py --input spec.json --dry-run
# Via ClawBio runner
python clawbio.py run skill-builder --demo
python clawbio.py run skill-builder --input spec.json
Demo
python clawbio.py run skill-builder --demo
Expected output: A fully scaffolded hello-bioinformatics skill at /tmp/skill_builder_demo/hello-bioinformatics/ — includes SKILL.md, hello_bioinformatics.py, tests/test_hello_bioinformatics.py, and a result.json + report.md in the skill-builder output directory documenting what was created.
Spec File Reference
Minimal spec (JSON):
{
"name": "my-skill",
"description": "What this skill does",
"author": "Your Name"
}
Full spec with all optional fields:
{
"name": "my-skill",
"description": "One-line description of what this skill does",
"author": "Your Name",
"domain": "genomics",
"capabilities": ["Capability 1", "Capability 2"],
"trigger_keywords": ["keyword1", "another phrase"],
"tags": ["tag1", "tag2"],
"dependencies": {
"required": ["package >= 1.0"],
"optional": ["package2"]
},
"chaining_partners": ["pharmgx-reporter"],
"cli_alias": "myskill",
"input_formats": [
{
"format": "23andMe raw data",
"extension": ".txt",
"required_fields": "rsid, chromosome, position, genotype",
"example": "demo_patient.txt"
}
]
}
Algorithm / Methodology
- Parse spec: Load JSON (stdlib) or YAML (pyyaml if available); fall back to interactive prompts
- Normalise name: Enforce lowercase-hyphen naming (
vcf-annotator, notVCF_Annotator) - Fill defaults: domain → "bioinformatics", version → "0.1.0", capabilities/triggers → generic placeholders
- Render SKILL.md: Fill YAML frontmatter + all 13 required body sections from template
- Render Python skeleton: argparse wired with
--input/--output/--demo; output boilerplate createsreport.md,result.json, reproducibility bundle - Render test skeleton: pytest fixture + 3 standard tests (demo runs, report generated, result.json valid)
- Validate: Run the 13-item CONTRIBUTING checklist against the generated SKILL.md before writing
- Register: Append catalog entry; patch
clawbio.pySKILLS dict via targeted string replacement
Example Queries
- "Create a new skill called vcf-annotator that annotates VCF files with ClinVar"
- "Scaffold a skill for running PLINK GWAS pipelines"
- "Build a skill template for GO enrichment analysis"
- "Validate my SKILL.md before I submit a PR"
Output Structure
output_directory/
├── report.md # Summary of what was generated
├── result.json # Machine-readable scaffold manifest
└── reproducibility/
└── commands.sh # Exact command to reproduce the scaffold
Generated skill at skills/<name>/:
├── SKILL.md # Complete skill definition
├── <name>.py # Python skeleton with --input/--output/--demo
├── tests/
│ └── test_<name>.py # pytest skeleton with 3 standard tests
└── examples/
└── example_spec.json # The spec that generated this skill
Dependencies
Required (stdlib only — zero install):
- Python 3.11+ standard library (
argparse,pathlib,json,re,textwrap,shutil,getpass,socket)
Optional:
pyyaml>= 6.0 — enables YAML spec files in addition to JSON; graceful fallback to JSON-only mode if absent
Safety
- Local-first: No network calls; all generation is offline
- Non-destructive: Never overwrites existing files without
--force; prompts or errors if destination exists - No hallucinated science: All generated SKILL.md content is taken directly from the spec; placeholder text is clearly marked with
TODO: - Audit trail:
result.jsonandcommands.shrecord exactly what was generated and when
Integration with Bio Orchestrator
Trigger conditions — the orchestrator routes here when:
- User says "create a skill", "scaffold a skill", "new skill", "build a skill", "add a skill"
- User provides a JSON/YAML file with
name,description,authorfields and asks to build a skill
Chaining partners:
bio-orchestrator: Skill builder output feeds back into the orchestrator once registered
Citations
- CONTRIBUTING.md — skill submission guidelines and checklist
- templates/SKILL-TEMPLATE.md — canonical SKILL.md template
04 同仓库更多 Skills
zarr-python
Stores and queries chunked N-D scientific arrays with Zarr-Python 3, including codecs, sharding, S3/GCS storage, and NumPy/Dask/Xarray integration. Use for array layout, bounded I/O, format migration, or scientific metadata preservation.
yeet
Publish local changes to GitHub by confirming scope, committing intentionally, pushing the branch, and opening a draft PR through the GitHub app from this plugin, with `gh` used only as a fallback where connector coverage is insufficient.
xarray
N-dimensional labeled arrays and datasets in Python. Built on top of NumPy and Dask. It introduces labels in the form of dimensions, coordinates, and attributes on top of raw NumPy-like arrays, making data analysis in physical sciences more intuitive and less error-prone. Use for working with multi-dimensional scientific data, NetCDF/GRIB/Zarr files, climate/weather/oceanographic datasets, remote sensing, geospatial imaging, large out-of-memory datasets with Dask, and labeled array operations.
writing-mstest-tests
Use when asked to fix, rewrite, update, improve, modernize, show corrected code for, or explain existing MSTest tests or MSTest-specific configuration. Use for "review" when corrected code or edits are wanted, even for one pasted assertion or passing tests with bad failure output. Covers expected/actual labels; generic Boolean, collection, string, numeric, null, identity, exception, hard-cast, and object[] checks; TestContext/lifecycle; timeout/cancellation; OS/CI conditions, retry, cleanup, parallelization, MSTest.Sdk project setup, and MSTESTxxxx. Honor the installed MSTest version. DO NOT USE to design new test cases (code-testing), perform report-only audits, create project files rather than explain MSTest setup, run tests, migrate frameworks, or handle non-MSTest/non-.NET code.
wgs-prs
End-to-end WGS to polygenic risk score pipeline. Takes paired-end FASTQ files (or a pre-existing VCF) through nf-core/sarek for variant calling, applies VCF QC (normalisation, hard filtering, Ti/Tv and Het/Hom checks), then computes polygenic risk scores via the PGS Catalog. Fills the FASTQ to VCF gap upstream of the gwas-prs skill.
webapp-testing
Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.