Initial commit

This commit is contained in:
Zhongwei Li
2025-11-30 08:57:35 +08:00
commit 6aae23c1d2
19 changed files with 4572 additions and 0 deletions

43
commands/list-skills.md Normal file
View File

@@ -0,0 +1,43 @@
---
description: List all available Claude Code skills in a formatted table
allowed-tools:
- Bash
- Read
- Glob
---
List all available Claude Code skills and display them in a readable format optimized for terminal viewing.
**Group skills by plugin source** to keep related skills together. Use this format:
```
Available Skills:
Plugin: {plugin-source}
{skill-name}
Description: {description}
{skill-name-2}
Description: {description}
Plugin: {another-plugin-source}
{skill-name-3}
Description: {description}
```
**Important formatting rules:**
- Group all skills from the same plugin/marketplace together
- Show "Plugin: {source}" header once per group
- Under each skill, show ONLY the skill name and description (don't repeat the plugin source)
- If a skill appears multiple times (e.g., "2 instances"), note that in the skill name line
- Keep full descriptions (no truncation) - natural line wrapping is fine
- Add blank lines between plugin groups for visual separation
At the end, include usage instructions:
```
Usage:
Invoke with: Skill tool using skill name
Example: ui-engineering:senior-ui-engineer
```

View File

@@ -0,0 +1,65 @@
---
description: Review a Claude Code slash command and recommend improvements based on best practices
argument-hint: [path-to-command.md]
allowed-tools:
- Read
- WebFetch
- Glob
---
Review the Claude Code slash command located at: **$1**
First, fetch the official documentation from https://docs.claude.com/en/docs/claude-code/slash-commands.md to understand current best practices.
Then read the command file and analyze it for conformance to best practices. Provide actionable recommendations for improvements.
Focus on:
- **Frontmatter Quality**:
- Presence of `description` field (clear, concise explanation)
- Proper use of `argument-hint` for auto-completion guidance
- `allowed-tools` specification (explicit permissions vs. inherited)
- `model` specification if needed for specific Claude models
- `disable-model-invocation` if command shouldn't be auto-invoked
- **Argument Handling**:
- Appropriate use of `$1`, `$2` positional parameters vs `$ARGUMENTS`
- Clear documentation of expected arguments in the prompt
- Sensible defaults for optional arguments
- Validation logic for required arguments
- **Command Focus**:
- Single, well-defined purpose (not trying to do too much)
- Clear use case that's distinct from skills/agents
- Appropriate for "simple prompt snippets used often"
- **Writing Style**:
- Clear, actionable instructions
- Proper use of file references with '@' prefix if needed (e.g., @path/to/file.md)
- Proper use of bash execution with '!' prefix if needed
- Examples provided where helpful
- **Tool Permissions**:
- 'allowed-tools' explicitly declared in frontmatter
- Only necessary tools included
- Bash commands specific rather than broad if using '!' prefix
- **Documentation Quality**:
- Description field matches actual command behavior
- Argument expectations clearly documented
- Usage examples provided where helpful
Provide prioritized, specific recommendations organized by:
1. **Critical** - Must fix for proper functionality or security
2. **Important** - Should fix for best practices alignment
3. **Nice-to-have** - Optional improvements for enhanced quality
For each recommendation:
- Explain the issue clearly with reference to official documentation
- Provide concrete examples of how to fix it
- Reference specific lines in the command file when applicable
- Show before/after examples for clarity
Highlight what the command does well and offer to implement improvements if requested.
If the command file path is not provided, search for `.md` files in `.claude/commands/` in the current working directory and offer to review them.

39
commands/review-skill.md Normal file
View File

@@ -0,0 +1,39 @@
---
description: Review a Claude Code skill and recommend improvements to conform to best practices
argument-hint: [path/to/skill/directory or skill-name]
allowed-tools:
- Task
- Skill
- Read
- Glob
---
Review the Claude Code skill located at: **$1**
Use the Skill tool to invoke `example-skills:skill-creator` to access comprehensive skill creation guidance, validation scripts, and packaging tools.
Read all files in the skill directory including SKILL.md, reference materials, templates, and scripts. Review the skill for conformance to best practices and provide actionable recommendations for improvements.
Focus on:
- **Metadata Quality**: Frontmatter completeness (name, description fields) with clear activation triggers and technology keywords
- **Progressive Disclosure**: Proper use of bundled resources (scripts/, references/, assets/) and when to load them
- **Writing Style**: Imperative/infinitive form (verb-first instructions), not second person
- **Prompt Quality**: Clarity, structure, effectiveness, and adherence to skill creation process
- **Examples and Usage**: Clear guidance on when to use and when NOT to use
- **Tool Specifications**: Proper tool usage patterns and integrations
- **Validation**: Run validation checks if available (mention `scripts/package_skill.py` for validation)
Provide prioritized, specific recommendations organized by:
1. **Critical** - Must fix for proper functionality
2. **Important** - Should fix for best practices alignment
3. **Nice-to-have** - Optional improvements for enhanced quality
For each recommendation:
- Explain the issue clearly
- Provide concrete examples of how to fix it
- Reference specific files and line numbers when applicable
Highlight what the skill does well and offer to implement improvements if requested. If the skill appears ready for distribution, mention that `scripts/package_skill.py` can validate and package it.
If the skill directory path is not provided, search for skill directories in `.claude/skills/` or `plugins/*/skills/` in the current working directory and offer to review them.