๐ŸŽฏ

skill-builder

๐ŸŽฏSkill

from jaganpro/sf-skills

VibeIndex|
What it does

Generates interactive wizard for creating, validating, and configuring Claude Code skills with comprehensive step-by-step guidance and best practices.

๐Ÿ“ฆ

Part of

jaganpro/sf-skills(34 items)

skill-builder

Installation

Install ScriptRun install script
curl -sSL https://raw.githubusercontent.com/Jaganpro/sf-skills/main/tools/install.sh | bash
git cloneClone repository
git clone https://github.com/Jaganpro/sf-skills
PythonRun Python server
python tools/installer.py --cli opencode --all # OpenCode
PythonRun Python server
python tools/installer.py --cli codex --all # Codex (OpenAI)
PythonRun Python server
python tools/installer.py --cli gemini --all # Gemini (Google)

+ 17 more commands

๐Ÿ“– Extracted from docs: jaganpro/sf-skills
8Installs
-
AddedFeb 4, 2026

Skill Details

SKILL.md

>

Overview

# Skill-Builder: Claude Code Skill Creation Wizard

Expert skill architect for Claude Code. Help users create well-structured, validated skills through an interactive wizard process.

Core Responsibilities

  1. Interactive Skill Creation: Step-by-step wizard to create new skills
  2. Template Application: Apply minimal-starter template with custom metadata
  3. Deep Validation: Validate YAML frontmatter, tool permissions, structure
  4. Bulk Validation: Validate all installed skills with comprehensive reporting
  5. Interactive Editor (v2.0): Terminal-based editor for refining skills
  6. Dependency Management (v2.0): Check, validate, visualize skill dependencies
  7. Testing Support: Verify skills work after creation
  8. Best Practices: Educate on skill design patterns

Skill Creation Workflow

Phase 1: Information Gathering

Use AskUserQuestion to collect (in sequence):

| # | Question | Format | Validation |

|---|----------|--------|------------|

| 1 | Skill name | Text | kebab-case required |

| 2 | Description | Text | One clear sentence |

| 3 | Author | Text | Optional |

| 4 | Tools needed | Multi-select | From valid tools list |

| 5 | Optional components | Multi-select | README, examples/, templates/, scripts/, docs/ |

| 6 | Tags | Text | Comma-separated |

| 7 | Location | Choice | Global (~/.claude/skills/) or Project (.claude/skills/) |

Valid tools: Bash, Read, Write, Edit, Glob, Grep, WebFetch, AskUserQuestion, TodoWrite, SlashCommand, Skill, BashOutput, KillShell

Phase 2: Scaffolding

  1. Determine path: Global โ†’ ~/.claude/skills/{name}/ | Project โ†’ .claude/skills/{name}/
  2. Check existing: If exists, ask to overwrite
  3. Create structure:

```bash

mkdir -p {base_path}

# Create optional dirs based on selection: examples/, templates/, scripts/, docs/

```

  1. Load template:

- Marketplace: Read: ~/.claude/plugins/marketplaces/sf-skills/skill-builder/templates/minimal-starter.md

- Project: Read: [project-root]/skill-builder/templates/minimal-starter.md

  1. Replace placeholders: {{SKILL_NAME}}, {{SKILL_DESCRIPTION}}, {{VERSION}}, {{AUTHOR}}, {{TAGS}}, {{ALLOWED_TOOLS}}
  2. Write files: SKILL.md + optional README.md, examples/example-usage.md

Phase 3: Deep Validation

```bash

python3 ${CLAUDE_PLUGIN_ROOT}/scripts/validate_yaml.py {base_path}/SKILL.md

```

Validation checks:

  • YAML syntax valid
  • Required fields: name, description, version
  • Name is kebab-case
  • Version is semver (X.Y.Z)
  • All tools in allowed-tools are valid (case-sensitive)
  • SKILL.md has content after frontmatter

If validation fails: Report errors with line numbers and fixes. See docs/validation-errors.md (in skill-builder folder) for detailed examples.

Phase 4: Testing (Optional)

Ask: "Test this skill now?" โ†’ Explain restart required, provide invocation example.

Phase 5: Completion Summary

```

โœ“ Skill '{name}' created successfully!

๐Ÿ“ Location: {full_path}

๐Ÿ“„ Files: SKILL.md [+ README.md, examples/, ...]

Next Steps:

  1. Customize SKILL.md with your logic
  2. Restart Claude Code to load skill
  3. Test: "Use the {name} skill to [task]"

Resources: docs/skill-structure.md, docs/frontmatter-reference.md, docs/best-practices.md (in skill-builder folder)

```

Validation Errors

| Error | Fix |

|-------|-----|

| YAML syntax | Quote strings with : or # |

| Missing field | Add required field to frontmatter |

| Invalid tool | Use exact case: Bash not bash |

| Bad version | Use X.Y.Z format |

| Bad name | Use kebab-case |

See docs/validation-errors.md (in skill-builder folder) for details.

Bulk Validation (v2.0)

When user asks to validate all skills:

```bash

python3 ${CLAUDE_PLUGIN_ROOT}/scripts/bulk_validate.py [--errors-only] [--format json]

```

Interpret results: Total skills, valid count, warnings, errors.

Guide fixes: Use interactive editor or manual edits.

See docs/workflow-examples.md (in skill-builder folder) for output examples.

Interactive Editor (v2.0)

For editing existing skills:

```bash

python3 ${CLAUDE_PLUGIN_ROOT}/scripts/interactive_editor.py /path/to/skill/

```

Features: Real-time validation, field editing, tool management, preview changes, auto-backup.

Commands: [e] Edit field | [t] Manage tools | [v] Validate | [s] Save | [r] Reload | [q] Quit

See [docs/workflow-examples.md](docs/workflow-examples.md) for UI example.

Dependency Management (v2.0)

Dependency syntax in SKILL.md:

```yaml

dependencies:

- name: other-skill

version: ">=1.2.0" # or ^1.2.0, ~1.2.0, *, exact

required: true

```

Version constraints: ^ (compatible) | ~ (approximate) | >= | exact | * (any)

Commands:

```bash

python3 ${CLAUDE_PLUGIN_ROOT}/scripts/dependency_manager.py check my-skill # Check deps

python3 ${CLAUDE_PLUGIN_ROOT}/scripts/dependency_manager.py tree my-skill # Visualize tree

python3 ${CLAUDE_PLUGIN_ROOT}/scripts/dependency_manager.py circular my-skill # Detect cycles

python3 ${CLAUDE_PLUGIN_ROOT}/scripts/dependency_manager.py validate --all # Validate all

```

See docs/workflow-examples.md (in skill-builder folder) for output examples.

Patterns & Best Practices

Skill types: Code analysis (Globโ†’Readโ†’Grepโ†’Report) | Documentation (Readโ†’Write) | Interactive (AskUserQuestionโ†’Validateโ†’Execute) | Testing (Bashโ†’Analyzeโ†’Report)

Best practices: Single responsibility | Clear kebab-case names | Minimal tools | Include examples | Proper semver

Troubleshooting

| Issue | Check |

|-------|-------|

| Creation fails | Permissions on ~/.claude/skills/, Python 3, disk space |

| Validation fails | YAML syntax, tool case-sensitivity (Bash not bash) |

| Skill doesn't load | Restart Claude Code, verify path, check frontmatter |

Notes

  • Restart Claude Code after creating skills
  • Test with real scenarios
  • This meta-skill can create other meta-skills!

---

License

MIT License. See [LICENSE](LICENSE) file.

Copyright (c) 2024-2025 Jag Valaiyapathy