command-builder（command 构建器）

[H1][COMMAND-BUILDER]

Dictum: *Reusable prompt templates reduce repetition and enforce consistency.*

Load commands from .claude/commands/ (project) or ~/.claude/commands/ (user). Filename becomes command name.

Tasks:

1. Read index.md — Reference file listing for navigation
2. Read workflow.md — 5-phase creation process
3. Read variables.md — Argument capture, file references, skill loading
4. Read hints.md — Argument-hint syntax, enum patterns
5. (prose) Load style-standards skill — Voice, formatting, constraints
6. Execute per workflow — UNDERSTAND, ACQUIRE, RESEARCH, AUTHOR, VALIDATE
7. Validate — Quality gate; see §VALIDATION

[REFERENCE]: index.md — Complete reference file listing

[1][FRONTMATTER]

Dictum: *Declarative constraints gate command execution.*

---
description: Brief description shown in /help menu
argument-hint: [required-arg] [optional-arg?]
allowed-tools: Read, Write, Edit, Glob, Grep, Task, Bash, TodoWrite
model: opus
---

[IMPORTANT]:

• [ALWAYS] Use --- delimiters (three dashes, no spaces).
• [ALWAYS] Quote strings containing :, #, [, ], {, }.
• [NEVER] Use tabs; YAML requires spaces.

[CRITICAL]:

• [ALWAYS] Declare Read for every @path reference.
• [ALWAYS] Declare Bash for every !command reference.
• [NEVER] Omit required tools—command fails without output.

[2][MODELS]

Dictum: *Model selection balances capability against cost and latency.*

[CRITICAL] Session inherits default model. Override for specific capability requirements.

[3][VARIABLES]

Dictum: *Dynamic substitution enables reusable commands.*

[CRITICAL]:

• [NEVER] Mix $ARGUMENTS and $1-$N in same command.
• [ALWAYS] Declare required tools for @path and !command.

→references/variables.md: Complete reference—examples, skill loading, anti-patterns.

[4][PATTERNS]

Dictum: *Canonical patterns accelerate development.*

[4.1][FILE_ANALYSIS]

---
description: Analyze file for issues
argument-hint: [file-path]
allowed-tools: Read
---
## Target
@$1
## Task
Identify security, performance, and code quality issues.

[4.2][MULTI_FILE_OPERATION]

---
description: Process files matching pattern
argument-hint: [glob-pattern]
allowed-tools: Read, Edit, Glob, TodoWrite
---
## Task
Match files via $1. Analyze content. Apply fixes. Track progress via TodoWrite.

[4.3][AGENT_WORKFLOW]

---
description: Multi-agent analysis
argument-hint: [target-folder]
allowed-tools: Task, Read, Glob, TodoWrite
---
## Task
1. Match files in $1 via Glob
2. Spawn analysis agents via Task
3. Synthesize findings, track via TodoWrite

[4.4][SKILL_CONTEXT]

---
description: Validate target against skill standards
argument-hint: [target] [focus?]
allowed-tools: Read, Task, Glob, Edit, TodoWrite
---
## Skill Context
@.claude/skills/[skill-name]/SKILL.md
@.claude/skills/[skill-name]/references/[domain]/*.md
## Task
Load context above. Spawn Task agents. Verify findings against loaded context. Apply corrections.

→templates/command-template.md — canonical template.

[5][ORGANIZATION]

Dictum: *Namespaces prevent command collision.*

.claude/commands/
├── git/
│   ├── commit.md  -> /git:commit
│   └── pr.md      -> /git:pr
└── test/
    └── unit.md    -> /test:unit

[6][VALIDATION]

Dictum: *Validation gates prevent incomplete artifacts.*

[VERIFY] Completion:

• Workflow: All 5 phases executed (UNDERSTAND → VALIDATE).
• Frontmatter: Valid YAML, description present, tools declared.
• Variables: No $ARGUMENTS + $1-$N mixing.
• Tools: All @path have Read, all !command have Bash.
• Quality: LOC < 125, verb-first naming.

[REFERENCE] Operational checklist: →validation.md