skill-factory-pipeline
por martinforsulu
Coding Agents & IDEsResumen
Multi-Agent Pipeline Orchestrator that builds new skills from scratch.
Descripción
Qué hace esta skill
A 7-stage multi-agent pipeline that turns a raw skill idea into a complete, market-ready skill package. Each stage runs as an isolated, sequential CLI agent call with file-based state between stages. Any stage can be re-run independently if it fails.
The sequential, file-based architecture means each stage is inspectable and re-runnable without re-executing the entire pipeline from scratch.
Cuándo usarla
- Turning a vague skill idea into a structured, deployable package
- Running market research before committing to building a skill
- Resuming a failed pipeline from a specific stage without restarting
- Generating pricing and positioning alongside the skill artifact
- Auditing and reviewing skill documentation before publishing
Flujo de trabajo de ejemplo
User asks: build a skill that automates changelog generation from git commits
- 1
Run init_pipeline.py with the idea string to create a workspace and idea.md
- 2
Review and edit idea.md to refine scope
- 3
Run pipeline.sh to execute all 7 agents sequentially
- 4
Monitor stage gates; if auditor fails, fix the issue and re-run from --from auditor
- 5
Collect workspace outputs: market.md, plan.md, arch.md, skill/, audit.md, docs_review.md, pricing.md
A complete skill package with market research, architecture, built skill files, audit report, documentation review, and pricing recommendation
Instalación
clawhub install martinforsulu/skill-factory-pipelineEjecuta el comando de la CLI o pega el prompt en Claude Code, Cursor, Codex o el agente de programación que uses.