From prompt-smith
Review Claude Code prompts (Skill/Agent/Command/Plugin) for quality and best practices
How this command is triggered — by the user, by Claude, or both
Slash command
/prompt-smith:review-prompt @path/to/file.md or @path/to/directory/ [--mode single|batch|consistency]Files this command reads when invoked
This command is limited to the following tools:
The summary Claude sees in its command listing — used to decide when to auto-load this command
# Prompt Review Command Review Claude Code prompts for quality, structure, and best practices compliance. ## Input **Target:** $ARGUMENTS ## Review Modes - **single** (default): Detailed review of a single file - **batch**: Summary + individual reports for multiple files - **consistency**: Check naming, structure, and style consistency across files ## Process ### Step 1: Determine Target and Mode Parse $ARGUMENTS to identify: - Target type: single file, directory, or glob pattern - Review mode: single, batch, or consistency (default: single for one file, batch for multiple) ### Ste...
Review Claude Code prompts for quality, structure, and best practices compliance.
Target: $ARGUMENTS
Parse $ARGUMENTS to identify:
For each file/directory, detect category from content:
.claude-plugin/plugin.jsonversion:, context:, effort: in frontmattermodel:, color:, or tools: in frontmatter, description starts with "Use this agent when", contains <example> blocks; may have memory:, background:, isolation: fieldsdescription: and optionally argument-hint: or allowed-tools:; no <example> blocksname (recommended, kebab-case, max 64 chars) and description (recommended)disable-model-invocation: true set for side-effect workflowscontext: fork skills have explicit instructions (not just guidelines)name (required) and description (required)<example> blocks presentcolor semantically matches purpose (recommended, universally used)hooks, mcpServers, or permissionModeWhen target is a plugin directory, review the overall plugin structure:
Structure (Critical)
.claude-plugin/plugin.json existsplugin.json has valid name (kebab-case, no spaces)commands/, agents/, skills/) are at plugin root, not inside .claude-plugin/.claude-plugin/ contains only plugin.jsonConfiguration (Major)
version follows semantic versioningdescription clearly describes plugin purposeplugin.json are relative (./ prefix)${CLAUDE_PLUGIN_ROOT}Security (Critical)
hooks, mcpServers, or permissionMode fieldsQuality (Minor)
claude plugin validate would pass (check structure)Components (Major)
# Prompt Review Result
## Summary
- **File**: [path]
- **Category**: [Skill/Agent/Command/Plugin]
- **Rating**: [A/B/C/D]
- **Critical Issues**: [count]
## Strengths
1. [strength 1]
2. [strength 2]
## Issues
### Critical (Must Fix)
1. **[Issue Title]**
- Current: [current state]
- Problem: [why it's a problem]
- Fix: [specific fix]
### Minor (Recommended)
1. **[Issue Title]**
- Current: [current state]
- Suggestion: [improvement]
## Corrected Sample
[Show corrected version of problematic sections]
## Checklist Results
| Item | Status | Comment |
|------|--------|---------|
| ... | OK/NG | ... |
# Batch Review Results
## Summary
| File | Category | Rating | Critical | Minor |
|------|----------|--------|----------|-------|
| ... | ... | ... | ... | ... |
## Statistics
- **Total Files**: [count]
- **Average Rating**: [A/B/C/D]
- **Total Critical Issues**: [count]
## Common Patterns
1. **[Pattern]**: [description] - Files: [list]
---
## Individual Reports
[Single mode format for each file]
# Consistency Review Results
## Target Directory
`[path]`
## Detected Categories
- Plugins: [count] directories
- Skills: [count] files
- Agents: [count] files
- Commands: [count] files
## Consistency Check
### Naming Conventions
| Item | Status | Details |
|------|--------|---------|
| File naming | OK/NG | [details] |
| Frontmatter name | OK/NG | [details] |
### Structure Consistency
| Item | Status | Details |
|------|--------|---------|
| Section structure | OK/NG | [details] |
| Example format | OK/NG | [details] |
### Style Consistency
| Item | Status | Details |
|------|--------|---------|
| description format | OK/NG | [details] |
| Body person | OK/NG | [details] |
## Recommended Actions
1. [action 1]
2. [action 2]
| Rating | Description |
|---|---|
| A | Meets all criteria, follows best practices |
| B | Minor issues only, no functional impact |
| C | Some important issues, but basically functional |
| D | Critical issues, needs major revision |
npx claudepluginhub nyoki/my-claude-plugins --plugin prompt-smith/review-promptScores user prompts against the RIOT framework (Role, Instructions, Output, Tone), identifies the primary failure mode, and generates an improved version with minimal changes.