Menu
Delegate specialized reviews to subagents and synthesize their findings. Use when reviewing code.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Lint prose against Simplified Technical English (ASD-STE100) rules. Use when the user asks to check, lint, or simplify technical writing (specs, manuals, procedures, docs) for STE compliance, or mentions ASD-STE100 or Simplified Technical English.
Any time committing, pushing, or creating new revisions is mentioned, use this commit flow.
Step-by-step procedure for building a Claude Code channel integration. Use when creating a new channel plugin that connects Claude Code to external systems via MCP, named pipes, webhooks, or chat platforms.
This skill should be used when the user asks to "set up a dev environment", "create a workspace", "open a browser pane", "split terminal", "spin up frontend and backend", "use cmux", "manage cmux panes", "run commands in cmux", or mentions cmux workspaces, panes, surfaces, or browser automation. Provides workflow guidance for orchestrating multi-pane development environments with terminal and browser surfaces.
Use when merging a stack of stacked PRs with jj. Triggered by phrases like "merge the PR stack", "merge the stacked PRs", "land these PRs", or "merge the stack into main".
Use when addressing PR review comments on stacked jj branches. Triggered by phrases like "address review comments", "fix PR feedback", "respond to review on stacked branch", or "update the PR with reviewer suggestions".
| name | Code Review Orchestrator |
| description | Delegate specialized reviews to subagents and synthesize their findings. Use when reviewing code. |
This skill orchestrates specialized review subagents to analyze code and documentation. Each subagent applies a specific review skill and returns focused findings. The orchestrator collects and presents results.
Determine scope:
Map content type to review skills:
For AI Prompt Content (skills, CLAUDE.md, commands):
prompt-brevity/SKILL.md - Token efficiency and clarityFor Python Files (skills, CLAUDE.md, commands):
python-code-review-types/SKILL.md - Ensuring python types are used correctly.For all Pull Request Files (changed files in PRs):
pr-file-review/SKILL.md - Identify unnecessary test artifacts and temporary files
IMPORTANT: If no specific review skill exists for the file type being reviewed, STOP and inform the user:
No specific review skill exists for [file type/category]. Available review skills:
- prompt-brevity (for AI prompts, skills, CLAUDE.md, slash commands)
To review [file type], create a new review skill at:
.claude/skills/collaboration/reviews/[skill-name].md
Use /create-skill to generate a new review skill template.
DO NOT perform any direct code review. Only delegate to existing review subskills.
For each selected review skill, launch a general-purpose subagent:
You are a specialized code reviewer. Apply the following skill to review the specified files.
**Skill to Apply**: [Read and follow .claude/skills/collaboration/reviews/[skill-name].md]
**Files to Review**: [list of file paths]
**Output Format**:
For each issue found, output:
---
**File**: [file path]
**Lines**: [line range, e.g., "45-52" inclusive or "78"]
**Issue**: [brief description]
**Suggestion**: [exact multiline replacement; omit if no specific change]
**Reason**: [why this improves the code]
---
Return ONLY the issues found. No commentary or summary.
Wait for all subagents to complete.
Output subagent results directly:
## Review Results
### Prompt Brevity Review
[paste subagent output]
### [Future Review Type]
[paste subagent output]
Do not add: Commentary, summaries, severity rankings, or additional suggestions.
Input: Review .claude/skills/testing/test-driven-development.md
Subagent Task:
You are a specialized code reviewer. Apply the following skill to review the specified files.
**Skill to Apply**: Read and follow .claude/skills/collaboration/reviews/prompt-brevity.md
**Files to Review**: .claude/skills/testing/test-driven-development.md
**Output Format**:
For each issue found, output:
---
**File**: [file path]
**Lines**: [line range inclusive]
**Issue**: [brief description]
**Suggestion**: [exact replacement; omit if no specific change]
**Reason**: [why this improves the code]
---
Return ONLY the issues found.
Orchestrator Output:
## Review Results
### Prompt Brevity Review
---
**File**: .claude/skills/testing/test-driven-development.md
**Lines**: 12-15
**Issue**: Verbose explanation with redundant phrasing
**Suggestion**: Test-driven development: write tests before implementation code
**Reason**: 68% token reduction (23 → 7 tokens) while preserving core message
---
---
**File**: .claude/skills/testing/test-driven-development.md
**Lines**: 45
**Issue**: Fluff phrase "you should make sure to"
**Suggestion**: run the tests
**Reason**: Direct imperative is clearer and saves 5 tokens
---
Input: Review all skills in .claude/skills/collaboration/
Action: Launch one subagent with all file paths
Orchestrator Output:
## Review Results
### Prompt Brevity Review
---
**File**: .claude/skills/collaboration/writing-commit-messages.md
**Lines**: 8-10
**Issue**: Ceremonial phrase "in order to"
**Suggestion**: to write good commit messages
**Reason**: Standard brevity pattern saves 2 tokens
---
---
**File**: .claude/skills/collaboration/writing-commit-messages.md
**Lines**: 67-72
**Issue**: Redundant examples demonstrating same concept
**Reason**: Reduces redundancy by 40 tokens without information loss (no specific replacement provided - requires context-aware consolidation)
---
---
**File**: .claude/skills/collaboration/code-review.md
**Lines**: 34
**Issue**: Hedge words "you might want to consider"
**Suggestion**: consider
**Reason**: Removes unnecessary hedging, saves 4 tokens
---
❌ Don't: Perform direct code reviews when no subskill exists
❌ Don't: Add your own commentary or summary
❌ Don't: Filter or editorialize subagent results
❌ Don't: Combine findings from different review types
❌ Don't: Launch subagents sequentially
❌ Don't: Add severity ratings or prioritization
To add new review dimensions:
.claude/skills/collaboration/reviews/[new-skill].mdExample future skills:
reviews/security-practices - Check for security anti-patternsreviews/accessibility - Verify inclusive language and examplesreviews/consistency - Ensure style consistency across skillscollaboration/reviews/prompt-brevity - Token efficiency review skillcollaboration/pr-file-review - Pull request file artifact reviewmeta/creating-skills - Skill creation guidelinesRemember: Orchestrate, don't review. Delegate to experts and present findings.