sync-skills-readme
Sync root README.md with current skills inventory from skills/ directory
mkdir -p ~/.claude/commands && curl -fsSL https://raw.githubusercontent.com/softaworks/agent-toolkit/HEAD/commands/sync-skills-readme.md -o ~/.claude/commands/sync-skills-readme.mdsync-skills-readme.md
# Sync Skills README Update the root README.md "Available Skills" table with all skills currently in the `skills/` directory. This command scans skills, reads their metadata, categorizes them, and regenerates the table. ## Workflow ### 1. Scan Skills Directory Use Bash to list all skill directories: ```bash ls -1 skills/ ``` For each skill directory found, proceed to step 2. ### 2. Extract Skill Metadata For each skill, read the SKILL.md frontmatter to extract: - `name:` field (skill identifier) - `description:` field (what the skill does) Use the Read tool to read `skills/[skill-name]/SKILL.md` and extract the YAML frontmatter between the `---` markers. **Example:** ```yaml --- name: codex description: Use when the user asks to run Codex CLI for code analysis... --- ``` Extract both fields for table generation. ### 3. Categorize Skills Assign each skill to a category based on its purpose. Use this mapping: | Category | Emoji | Keywords in name/description | |----------|-------|------------------------------| | AI Tools | 🤖 | codex, gemini, perplexity, ai, llm, model | | Meta | 🔮 | command-creator, plugin-forge, plugin, command | | Documentation | 📝 | docs, documentation, handoff, requirements, diagram, mermaid, draw, excalidraw, marp, slide, c4-architecture | | Development | 🛠️ | session, handoff, entropy, development, workflow, database, dependency | | Design & Frontend | 🎨 | design, frontend, ui, openapi, typescript, system | | Utilities | 🔧 | domain, meme, web-to-markdown, utility, tool, datadog | | Planning | 🎯 | plan, planning, spec, forge, gepetto, requirements, clarity, game-changing, features | | Professional | 👔 | professional, communication, career, soft-skill, feedback, conversation | | Testing | 🧪 | test, testing, qa, quality | | Git | 📦 | commit, git, branch, pr, pull-request | **Categorization logic:** 1. Check the skill name and description (case-insensitive) 2. Match against keywords in the table above 3. Assign to the first matching category 4. If no match, default to "Utilities" (🔧) ### 4. Generate Compact Table Create a markdown table with this exact format: ```markdown ## 📚 Available Skills | Category | Skill | Description | |----------|-------|-------------| | 🤖 AI Tools | [codex](skills/codex/README.md) | Advanced code analysis with GPT-5.2 | | 🤖 AI Tools | [gemini](skills/gemini/README.md) | Large-scale review (200k+ context) | ... ``` **Table structure:** - Column 1: Category emoji + name (e.g., "🤖 AI Tools") - Column 2: Skill name as markdown link to its README - Column 3: Short description (max ~50 chars, extracted from SKILL.md description field) **Sorting:** - Group by category (AI Tools first, then Meta, Documentation, Design & Frontend, Development, Planning, Professional, Testing, Git, Utilities last) - Within each category, sort skills alphabetically by name **Description shortening:** - If description is longer than 50 characters, create a shortened version - Keep it concise and action-oriented - Examples: - "Use when the user asks to run Codex CLI for code analysis, refactoring, or automated editing" → "Advanced code analysis with GPT-5.2" - "Create comprehensive API handoff documentation for frontend developers after backend implementation" → "API handoff docs for frontend" ### 5. Update README.md Use the Edit tool to replace the "Available Skills" section in the root README.md. **Find and replace:** 1. Locate the section starting with `## 📚 Available Skills` 2. Find the end of the table (marked by the next `---` separator or next `##` heading) 3. Replace everything between those markers with the newly generated table **Important:** - Preserve the `---` separator after the table - Do NOT modify other sections (Quick Navigation, Installation, etc.) - Only replace the table content, keep the section heading ### 6. Report Results After updating README.md, report to the user: ``` ✅ README.md updated successfully 📊 Summary: - Total skills: [count] - Categories: [list of categories with counts] - Skills added/updated: [list if any changes detected] The Available Skills table in README.md now reflects all skills in the skills/ directory. ``` ## Error Handling **If a skill directory doesn't have SKILL.md:** - Skip that skill - Warn the user: "⚠️ Skipped [skill-name]: No SKILL.md found" **If SKILL.md doesn't have required frontmatter:** - Skip that skill - Warn: "⚠️ Skipped [skill-name]: Missing name or description in frontmatter" **If README.md section not found:** - Error and stop: "❌ Could not find '## 📚 Available Skills' section in README.md" ## Tools to Use - **Bash** - List skills directory - **Read** - Read SKILL.md files for metadata extraction - **Edit** - Update README.md with new table - **DO NOT use Grep** - Read files directly for accurate frontmatter parsing ## Success Criteria - All valid skills from skills/ directory appear in the table - Skills are correctly categorized with appropriate emojis - Table maintains compact format (3 columns) - Descriptions are concise and readable - README.md is updated without breaking other sections
Refactor bloated AGENTS.md, CLAUDE.md, or similar agent instruction files to follow progressive disclosure principles. Splits monolithic files into organized, linked documentation.
Use this agent when you need to visualize UI concepts through ASCII mockups before implementation. Examples: <example>Context: User has an idea for a dashboard layout with data tables and charts. user: 'I want to create a dashboard that shows user analytics with a sidebar navigation, main content area with charts, and a data table below' assistant: 'I'll use the ascii-ui-mockup-generator agent to create multiple ASCII mockup variations for your dashboard concept.' <commentary>The user wants to visualize a UI concept, so use the ascii-ui-mockup-generator to create multiple ASCII representations they can choose from.</commentary></example> <example>Context: User is designing a form layout with multiple input fields. user: 'I need a contact form with name, email, message fields and a submit button' assistant: 'Let me use the ascii-ui-mockup-generator to create several ASCII mockup options for your contact form layout.' <commentary>Since the user needs to visualize form layouts, use the ascii-ui-mockup-generator to provide multiple ASCII design options.</commentary></example>
Create API handoff documentation for frontend developers. Use when backend work is complete and needs to be documented for frontend integration, or user says 'create handoff', 'document API', 'frontend handoff', or 'API documentation'.
Generate architecture documentation using C4 model Mermaid diagrams. Use when asked to create architecture diagrams, document system architecture, visualize software structure, create C4 diagrams, or generate context/container/component/deployment diagrams. Triggers include "architecture diagram", "C4 diagram", "system context", "container diagram", "component diagram", "deployment diagram", "document architecture", "visualize architecture".
codebase-pattern-finder is a useful subagent_type for finding similar implementations, usage examples, or existing patterns that can be modeled after. It will give you concrete code examples based on what you're looking for! It's sorta like codebase-locator, but it will not only tell you the location of files, it will also give you code details!
Use when the user asks to run Codex CLI (codex exec, codex resume) or references OpenAI Codex for code analysis, refactoring, or automated editing. Uses GPT-5.2 by default for state-of-the-art software engineering.
Create a detailed implementation plan using Codex 5.2 with high reasoning
This skill should be used when creating a Claude Code slash command. Use when users ask to "create a command", "make a slash command", "add a command", or want to document a workflow as a reusable command. Essential for creating optimized, agent-executable slash commands with proper structure and best practices.