agentsclimarketplace
Sign in

Skill builder

Skill Babalsaab/Project_Starter/.claude/skills/skill-builder

Use when creating new skills, optimizing existing skills, or auditing skill quality. Guides skill development following Claude Code official best practices and the Vibecoded Orchestrator's agent-skill architecture.From its SKILL.md

Install
npx -y skills add Babalsaab/Project_Starter --skill skill-builder

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

5.3 KB, as published. Nobody here has run it

What This Skill Does

Guides the creation and optimization of skills for both Claude Code (.claude/skills/) and the Vibecoded Orchestrator (skills/). Use this whenever:

  • Building a new skill from scratch
  • Optimizing or auditing an existing skill
  • Deciding on advanced features (subagent execution, hooks, dynamic context, etc.)
  • Troubleshooting a skill that isn't working correctly

For the complete technical reference on all frontmatter fields, advanced patterns, and troubleshooting, see reference.md.

Quick Start: Skill Types in This Project

This project has two skill systems:

SystemLocationFormatUsed By
Claude Code Skills.claude/skills/<name>/SKILL.mdYAML frontmatter + markdownClaude Code native /slash-commands
Orchestrator Skillsskills/<name>/SKILL.md + config.yamlYAML config + SKILL.md with frontmatterMCP server get_skill, list_skills tools

When to use which:

  • Claude Code skill — For interactive tasks invoked by /command or auto-detected by Claude. Runs in the AI coder session.
  • Orchestrator skill — For quality gates applied during pipeline execution. Used by agents via the MCP server.

Mode 1: Build a New Skill

Run the Discovery Interview first. Do NOT start writing files until discovery is complete.

Discovery Interview

Ask questions using AskUserQuestion, one round at a time.

Round 1: Goal & Name

  • What does this skill do? What problem does it solve?
  • What should we call it? (lowercase, hyphens, max 64 chars)
  • Which system? Claude Code skill (.claude/skills/) or Orchestrator skill (skills/)?

Round 2: Trigger

  • What would someone say to trigger this? (2-3 natural language phrases)
  • Should it be user-only (/slash-command), Claude-auto-invocable, or both?
  • Does it accept arguments? If so, what?

Round 3: Step-by-Step Process

  • Walk through exactly what happens from trigger to output
  • For each step: Does Claude do it directly, or delegate?
  • Conversational (back-and-forth) or fire-and-forget?

Round 4: Inputs, Outputs & Dependencies

  • What inputs does the skill need?
  • What does it produce? Where do outputs go?
  • External APIs, scripts, or tools needed?
  • Which orchestrator agents is this compatible with?

Round 5: Guardrails & Edge Cases

  • What could go wrong? Common failure modes?
  • What should this skill NOT do?
  • Cost concerns? (API calls, image generation, etc.)

Round 6: Confirmation Summarize understanding back to the user:

## Skill Summary: [name]

**Goal:** [one sentence]
**System:** Claude Code / Orchestrator / Both
**Trigger:** `/name` + [natural language phrases]
**Arguments:** [what it accepts, or "none"]

**Process:**
1. [step]
2. [step]

**Inputs:** [what it reads/needs]
**Outputs:** [what it produces + where]
**Compatible Agents:** [agent IDs if orchestrator skill]
**Guardrails:** [what can go wrong, what to avoid]

Build Phase

Step 1: Choose the target system

For Claude Code skills (.claude/skills/<name>/SKILL.md):

---
name: [gerund-name]
description: [3rd person, includes trigger keywords]
allowed-tools: [comma-separated tool list]
---

For Orchestrator skills (skills/<name>/), create both files:

  • config.yaml — id, name, description, version, domain, priority, compatible_agents
  • SKILL.md — YAML frontmatter + checklist content

Step 2: Write the content

  1. Context — Files to read, APIs to call
  2. Step-by-step workflow — Numbered steps with exact instructions
  3. Output format — Templates or examples
  4. Notes — Edge cases, constraints, what NOT to do

Step 3: Add supporting files (if needed)

  • Keep SKILL.md under 500 lines
  • Move detailed reference to supporting files
  • Reference them from SKILL.md

Step 4: Test

  • Natural language invocation
  • Direct /skill-name invocation
  • Edge cases with missing/unusual input

Mode 2: Audit an Existing Skill

Frontmatter Audit

  • name matches directory name
  • description uses natural trigger keywords
  • allowed-tools set if tool access should be restricted
  • context: fork used if self-contained
  • No unnecessary fields

Content Audit

  • Under 500 lines
  • Clear step-by-step workflow
  • Output format specified
  • All file paths documented
  • No vague instructions

Integration Audit

  • Documented in CLAUDE.md
  • Supporting files referenced (not orphaned)
  • Scripts have correct paths
  • API keys in env vars, never hardcoded

Quality Audit

  • A beginner could follow the instructions
  • Instructions are actionable, not abstract
  • Doesn't duplicate CLAUDE.md content
  • Output paths follow convention

Important Notes

  • Always read an existing skill before optimizing it
  • Check for similar skills that could be extended instead
  • For advanced patterns (subagent execution, hooks, permissions), see reference.md

Keep looking

Skills are one crate of 627,801. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.