command-development（命令开发）

Slash Commands Meta Skill

## 🚨 MANDATORY: Invoke docs-management First STOP - Before providing ANY response about slash commands: 1. INVOKE docs-management skill 2. QUERY for the user's specific topic 3. BASE all responses EXCLUSIVELY on official documentation loaded Skipping this step results in outdated or incorrect information. ### Verification Checkpoint Before responding, verify: - [ ] Did I invoke docs-management skill? - [ ] Did official documentation load? - [ ] Is my response based EXCLUSIVELY on official docs? If ANY checkbox is unchecked, STOP and invoke docs-management first.

Overview

Central authority for Claude Code slash commands. This skill uses 100% delegation to docs-management - it contains NO duplicated official documentation.

Architecture: Pure delegation with keyword registry. All official documentation is accessed via docs-management skill queries.

When to Use This Skill

Keywords: slash commands, custom commands, built-in commands, project commands, personal commands, plugin commands, MCP commands, Skill tool, command frontmatter, allowed-tools, argument-hint, $ARGUMENTS, bash execution, file references, namespacing, skills vs commands

Use this skill when:

• Creating custom slash commands
• Understanding built-in commands (/init, /compact, /memory, etc.)
• Configuring command frontmatter (allowed-tools, description, model)
• Using arguments in commands ($ARGUMENTS, $1, $2)
• Executing bash commands in slash commands
• Referencing files in commands (@file syntax)
• Understanding namespacing and organization
• Working with plugin commands
• Using MCP prompts as slash commands
• Configuring Skill tool permissions for commands
• Deciding between skills vs slash commands
• Troubleshooting command issues

Keyword Registry for docs-management Queries

Use these keywords when querying docs-management skill for official documentation:

Core Concepts

Command Types

Configuration

Features

Skill Tool (Unified)

Note: The SlashCommand tool has been unified into the Skill tool. Both slash commands and agent skills are now invoked via the single Skill tool. Update any existing permission rules from SlashCommand to Skill.

Comparisons

Quick Decision Tree

What do you want to do?

1. Create a custom command -> Query docs-management: "custom slash commands", "command creation"
2. Configure command frontmatter -> Query docs-management: "command frontmatter", "allowed-tools"
3. Use arguments in commands -> Query docs-management: "$ARGUMENTS", "command arguments"
4. Execute bash in commands -> Query docs-management: "bash command execution", "! prefix"
5. Reference files in commands -> Query docs-management: "file references commands", "@ prefix"
6. Understand namespacing -> Query docs-management: "command namespacing", "subdirectory commands"
7. Work with plugin commands -> Query docs-management: "plugin commands", "plugin slash commands"
8. Use MCP as commands -> Query docs-management: "MCP slash commands", "mcp__server__prompt"
9. Configure Skill tool for commands -> Query docs-management: "Skill tool", "Skill tool permissions"
10. Choose skills vs commands -> Query docs-management: "skills vs slash commands"
11. View built-in commands -> Query docs-management: "built-in slash commands table"
12. Troubleshoot command issues -> Query docs-management: "slash commands troubleshooting" + specific issue
13. Configure MCP command permissions -> Query docs-management: "MCP permissions wildcards", "mcp__servername"

Topic Coverage

Built-in Commands

• Complete command reference table
• Common commands (/init, /compact, /memory, /clear, /cost)
• Session management (/resume, /exit, /rewind)
• Configuration (/config, /permissions, /model)
• MCP and plugins (/mcp, /plugin, /agents)

Custom Commands

• Project commands (.claude/commands/)
• Personal commands (~/.claude/commands/)
• Command file format (markdown)
• Naming conventions (filename = command name)

Command Frontmatter

• allowed-tools: Tool restrictions
• argument-hint: Argument display hints
• description: Command description
• model: Model selection
• disable-model-invocation: Prevent Skill tool access

⚠️ YAML Parsing Warning (argument-hint): If your argument-hint contains colons inside brackets (e.g., [optional: text]), you MUST quote the value: ``yaml # WRONG - crashes with React error #31 argument-hint: [optional: target files] # CORRECT - quoted to force string interpretation argument-hint: "[optional: target files]" ` YAML interprets [key: value] as an array containing an object, which crashes Claude Code's React UI. Known issue: GitHub #16990, #16910. **Always quote argument-hint if it contains :** **Length Guidelines (argument-hint):** Keep argument-hints concise for readability in the Claude Code UI: | Length | Status | Action | | --- | --- | --- | | ≤60 chars | Ideal | No action needed | | 61-80 chars | Acceptable | Consider shortening | | 81-120 chars | Long | Review for clarity | | >120 chars | Too long | Must shorten | **Shortening Strategies:** - Use [flags] or [options] for many boolean flags - Use <type> placeholder instead of listing all a|b|c|d values - Omit rarely-used optional arguments - Group related options: [scope] instead of files|staged|pr **Examples:** `yaml # Too long (150+ chars) - lists every option argument-hint: [files or 'staged' or 'pr' or '--parallel' or '--sequential' or '--batched'] # Ideal (55 chars) - uses groupings argument-hint: "[scope] [--profile <name>] [flags]" ``

Arguments

• $ARGUMENTS: All arguments as single string
• $1, $2, $3...: Positional arguments
• Combining argument patterns
• Default values and handling

Bash Execution

• ! prefix for inline bash
• Including command output in context
• Required allowed-tools configuration
• Git commit example patterns

File References

• @ prefix for file inclusion
• Single and multiple file references
• Path resolution behavior

Namespacing

• Subdirectory organization
• (project) vs (user) indicators
• Conflict resolution rules
• Namespace display in /help

Plugin Commands

• Plugin command structure
• Namespaced invocation (/plugin-name:command)
• Plugin marketplace integration
• Command distribution

MCP Commands

• MCP prompt to command mapping
• /mcp__server__prompt pattern
• Dynamic discovery
• Authentication and permissions

Skill Tool Configuration (for Commands)

• Programmatic command execution via unified Skill tool
• Supported command requirements
• Permission rules (exact match, prefix match)
• Character budget limits
• Disabling tool access

Skills vs Slash Commands

• Complexity comparison
• Structure differences
• Discovery patterns (explicit vs automatic)
• When to use each approach

Delegation Patterns

Standard Query Pattern

User asks: "How do I create a custom command?"

1. Invoke docs-management skill
2. Use keywords: "custom slash commands", "command creation"
3. Load official documentation
4. Provide guidance based EXCLUSIVELY on official docs

Multi-Topic Query Pattern

User asks: "I want a command that runs git status and uses arguments"

1. Invoke docs-management skill with multiple queries:
   - "bash command execution", "! prefix commands"
   - "$ARGUMENTS", "command arguments"
   - "allowed-tools commands"
2. Synthesize guidance from official documentation

Comparison Query Pattern

User asks: "Should I use a skill or slash command for code review?"

1. Invoke docs-management skill
2. Use keywords: "skills vs slash commands", "when to use commands vs skills"
3. Guide user through decision criteria from official docs

Troubleshooting Quick Reference

Plugin Commands (claude-ecosystem)

This plugin provides slash commands in plugins/claude-ecosystem/commands/:

• /claude-ecosystem:scrape-docs - Scrape Claude documentation from official sources
• /claude-ecosystem:refresh-docs - Refresh local index without network scraping
• /claude-ecosystem:validate-docs - Validate index integrity and detect drift

Discovery:

• Use /help to see all available commands (built-in + plugin)
• Plugin commands can be invoked with or without the claude-ecosystem: prefix (when no conflicts)

Naming Conventions

Official Patterns (from Built-in Commands)

Claude Code's built-in commands use these patterns:

Key insight: Built-in commands do NOT use dot-namespacing or noun-first grouping.

Recommended Approach

For action commands (scrape, refresh, validate, create, etc.):

Use verb-noun kebab-case: /scrape-docs, /refresh-docs, /validate-docs

For display/status commands (built-in examples):

Use single noun: /status, /config, /todos

For simple utilities (built-in examples):

Use single verb or short descriptive name: /clear, /compact, /review

Plugin Command Namespacing

Plugin commands get automatic namespacing via /plugin-name:command:

• /claude-ecosystem:scrape-docs (fully qualified)
• /scrape-docs (short form when no conflicts)

This means you don't need to embed the domain in the command name—the plugin prefix handles disambiguation.

File Naming

Command files use the exact command name with .md extension:

• /scrape-docs -> commands/scrape-docs.md
• /refresh-docs -> commands/refresh-docs.md
• /validate-docs -> commands/validate-docs.md

Auditing Commands

This skill provides the validation criteria used by the command-auditor agent for formal audits.

Audit Resources

Scoring Categories

Thresholds: 85+ = PASS, 70-84 = PASS WITH WARNINGS, <70 = FAIL

Related Agent

The command-auditor agent (Haiku model) performs formal audits using this skill:

• Auto-loads this skill via skills: command-development
• Uses validation checklist and scoring rubric
• Generates structured audit reports
• Invoked by /audit-commands command

External Technology Validation

When auditing commands that use external technologies (scripts, packages, runtimes), the auditor MUST validate claims using MCP servers before flagging findings.

Technologies Requiring MCP Validation:

• .NET/C# scripts: Validate with microsoft-learn + perplexity
• Node.js/npm packages: Validate with context7 + perplexity
• Python scripts/packages: Validate with context7 + perplexity
• Shell scripts: Validate with perplexity
• Any version-specific claims: ALWAYS validate with perplexity

Validation Rule:

Never flag a technology usage as incorrect without first:

1. Querying appropriate MCP server(s) for current documentation
2. Verifying with perplexity for recent changes (especially .NET 10+)
3. Documenting MCP sources in the finding

Stale Data Warning:

• microsoft-learn can return cached/outdated documentation
• ALWAYS pair microsoft-learn with perplexity for version verification
• Trust perplexity for version numbers and recently-released features

References

Official Documentation (via docs-management skill):

• Primary: "slash-commands" documentation
• Related: "skills", "plugins", "MCP", "interactive-mode"

This Plugin:

• Plugin commands: plugins/claude-ecosystem/commands/
• Plugin skills: plugins/claude-ecosystem/skills/

Version History

• v1.0.5 (2026-01-10): Added YAML parsing warning for argument-hint

- Documented React error #31 caused by unquoted [key: value] in argument-hint - Added troubleshooting entry for React error #31 - Known issue: GitHub #16990, #16910

• v1.0.4 (2026-01-09): Updated for Claude Code v2.1.x SlashCommand → Skill tool unification

- Replaced all SlashCommand tool references with unified Skill tool - Added migration note explaining the unification - Updated keyword registry for new tool naming

• v1.0.3 (2025-11-30): Revised naming conventions based on official docs analysis

- Updated naming conventions to align with built-in command patterns - Recommended verb-noun kebab-case for action commands - Removed dot-namespacing recommendation (not used by built-in commands) - Documented plugin command automatic namespacing

• v1.0.2 (2025-11-27): Added Repository Naming Convention

- Added comprehensive naming convention section - Documented dot-namespacing for command families - Documented flat kebab-case for standalone commands - Added rationale and decision guidance - Added file naming examples

• v1.0.1 (2025-11-27): Audit and enhancement

- Added MCP permissions wildcards keywords - Added character budget keywords - Added MCP permissions decision path to quick decision tree - Added MCP permission denied troubleshooting entry

• v1.0.0 (2025-11-26): Initial release

- Pure delegation architecture - Comprehensive keyword registry - Quick decision tree - Topic coverage for all slash command features - Troubleshooting quick reference

Last Updated

Date: 2026-01-10 Model: claude-opus-4-5-20251101