skill-factory_skill

Overview

This skill is a research-backed workflow orchestrator that automates skill creation from discovery through final audit. It runs a 9-phase pipeline combining firecrawl-powered research gathering, structured synthesis, multi-tier validation, and a comprehensive audit against Anthropic specifications for Claude Code. The tool is designed to produce production-ready, high-quality skills with clear checkpoints and fail-fast handling.

How this skill works

The orchestrator detects whether you have existing research or need automated web scraping and then builds a TodoWrite task list for the chosen path. It invokes specialized commands sequentially—research, format, scaffold, write, content review, compliance checks, runtime and integration validation, and a final audit—waiting for each step to complete and applying one-shot fixes when possible. Dependency checks and a three-tier fix strategy ensure clear escalation and stop-on-complex-failure behavior.

When to use it

• Creating a new Claude Code skill from idea to validated artifact
• Automating research collection and formatting for a skill topic
• Enforcing content and technical quality before submitting a skill
• Running runtime and integration checks to prevent name conflicts or loading errors
• Needing a guided, repeatable workflow with checkpoints and auditability

Best practices

• Start with explicit research sources when available to speed up the workflow
• Run the full pipeline for new skills; use individual commands for targeted fixes
• Approve Tier 2 guided fixes promptly to avoid workflow stalls
• Do not run commands in parallel—wait for each step’s result before continuing
• Keep research artifacts by default for future iterations unless cleanup is needed

Example use cases

• Build an infrastructure skill documenting terraform best practices with automated scraping of official docs
• Create a developer helper skill that synthesizes coding patterns from multiple blog posts and examples
• Validate an existing skill after content edits to ensure runtime loading and naming uniqueness
• Run a compliance pass against Anthropic specs before opening a pull request
• Iteratively fix formatting and frontmatter issues using the tiered auto-fix strategy