bmad-workflow-builder

Builds, edits, and analyzes AI skills and workflows using outcome-driven quality standards.

Updated Sep 3, 2026
One-click install
npx skills add https://github.com/watchthelight/shatterfish --skill bmad-workflow-builder-watchthelight
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: bmad-workflow-builder
Source: https://github.com/watchthelight/shatterfish/tree/main/.claude/skills/bmad-workflow-builder
Command: npx skills add https://github.com/watchthelight/shatterfish --skill bmad-workflow-builder-watchthelight

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires tiktoken, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve? Turning a rough idea into a lean, effective AI skill is hard: most generated skills ship bloated instructions, generic procedure, and no evidence they beat the bare model. This Skill guides the full lifecycle of building, editing, and quality-analyzing skills so every line earns its place. ## Core Features & Use Cases - Guided skill building: Hardens a half-formed idea through conversation, grounds it in real project expertise, scaffolds the skill deterministically, and captures decisions in a resumable memlog. - Quality analysis lenses: Runs deterministic pre-pass scanners (token budgets, frontmatter validation, template artifacts, workflow integrity) plus LLM lenses, then renders a self-contained HTML report. - Eval and lint gates: Offers baseline and trigger eval modes, enforces SKILL.md token budgets, and runs lint scripts before handoff. - Use Case: Ask it to analyze an existing skill directory; it scans frontmatter, structure, and prompt quality, then produces a graded HTML report with severity-ranked findings. ## Quick Start Use the bmad-workflow-builder skill to analyze the skill at .claude/skills/my-skill and produce a quality report.

Frequently Asked Questions about bmad-workflow-builder

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build a new AI skill from an idea?

Invoke the skill with a description of what you want to build. It opens the floor for your goals and references, hardens the idea through conversation, scaffolds the skill with init_skill.py, and drafts a minimal outcome-driven SKILL.md.

How do I quality-check an existing skill?

Point the skill at the existing skill's path with words like analyze or quality check. It runs deterministic pre-pass scanners for token counts, frontmatter, and structure, then LLM lenses, and renders a graded HTML report.

What is the SKILL.md token budget enforced by the builder?

The default desired budget is 2000 tokens with a hard cap of 3000, measured with tiktoken cl100k_base. Skills over budget are fixed by lifting sections into references/ until they are back under the limit.

Does the builder support headless non-interactive runs?

Yes, pass --headless or -H to skip interactive questioning. Headless runs resume automatically from the memlog and emit a JSON status object with the skill path and completion state.

When should a skill include a customize.toml file?

Only when the author explicitly accepts end-user customization during the build; the default is no. Accepted skills get universal defaults like activation steps and persistent facts, merged across base, team, and personal override files.