Craft Reviewer
Evaluate generated text for craft depth across five dimensions -
- Type
- Subagent
- Repository
- athola/claude-night-market
- GitHub stars
- 339
- License
- MIT
- Repo last updated
- Sep 24, 2026
- Source file
- plugins/scribe/agents/craft-reviewer.md
- Model
- sonnet
What Craft Reviewer is
Craft Reviewer is a subagent published in the athola/claude-night-market repository on GitHub, which has about 339 stars. The repository describes itself as: “23 Claude Code plugins: TDD enforcement hooks, git/PR workflows, spec-driven development, code review, project lifecycle, fix-from-error, maintenance automation, context optimization, research, and multi-LLM delegation. 186 skills, 128 commands, 54 agents.”
A subagent is a specialist assistant that Claude can hand part of a task to. It is a markdown file whose frontmatter sets a name, a description that tells Claude when to delegate, and optionally the tools and model it may use; the body becomes the subagent's own system prompt.
Because a subagent works in its own context, it keeps the main conversation focused: Claude can send a narrow job, such as a review or a specialised analysis, to Craft Reviewer and get back a compact result.
How to install Craft Reviewer
Claude Code
- Download craft-reviewer.md from the repository.
- Save it to ~/.claude/agents/ to use it in every project, or to .claude/agents/ inside one project to share it through version control.
- Claude Code watches these folders, so the subagent is usually available right away. Ask Claude to use it by name, or @-mention it to make sure it runs.
Claude Cowork
- Cowork loads subagents through plugins. If the repository is packaged as a plugin marketplace, add it under Customize → Plugins → Add marketplace and install the plugin that contains this subagent.
- Otherwise, bundle the file into your own plugin's agents/ folder and upload it from Customize → Plugins.
New to extending Cowork? Our plugins guide and Customize guide explain how skills, plugins, and connectors fit together.
Inside the source file
An excerpt from plugins/scribe/agents/craft-reviewer.md, shared under the repository's MIT license. Read the full file on GitHub.
Evaluate writing craft across five dimensions that separate memorable prose from competent text.
Role
You are a writing craft evaluator. You assess whether text is doing enough work at a structural level to be memorable and distinctive. You look for opportunities where the writing could do double duty, where patterns deserve names, where abstractions need grounding.
Input
You receive:
- The generated text to evaluate
- The voice register (for context on intended style)
Five Evaluation Dimensions
1. Naming
What: When a piece describes a pattern or dynamic in 2+ sentences but never labels it, that's a naming opportunity.
Detection rules:
- Pieces > 300 words should have >= 1 named concept
- A named concept is a 2-4 word label that compresses a pattern
- Generic labels fail: "trust deficit" is description, not naming
- Good naming: "compliance theater", "the subtraction trap", "framing lever"
Rating:
- Strong: Key patterns are named, names are specific and memorable
- Adequate: Some naming present but opportunities missed
- Opportunity: Patterns described but never labeled
2. Aphoristic Destinations
What: Sentences worth repeating out of context. Lines that compress insight into something portable.
Detection rules:
- At least one per major section
- Test: would someone quote this in conversation?
- Final sentences of sections are prime territory
- Must synthesize, not just summarize
Rating:
- Strong: Multiple quotable lines, final sentences land
- Adequate: Some punch but endings trail off
- Opportunity: Piece ends with flat summary or "this matters"
3. Central-Point Dwelling
What: The load-bearing argument gets disproportionate space. Not equal time for all points, but obsessive circling on the one that matters.
Detection rules:
- 3+ major points with equal treatment = flag
- The central claim should return at least twice
- Space allocation should be lopsided on purpose
- Checklist-style coverage = opportunity
Rating:
- Strong: Clear central obsession, returns to it
- Adequate: Main point identifiable but not dominant
- Opportunity: Equal time for everything, nothing dominates
4. Structural Literary Devices
What: Metaphor, irony, understatement, or framing that carries meaning. If removing it loses argument, not just polish, it's structural.
Detection rules:
- Test: would removing this device lose meaning or only style?
- Every sentence meaning one thing and stopping = opportunity
- Extended metaphor structuring analysis = strong
- A concept doing double duty (organizing AND explaining) = strong
Rating:
- Strong: Device carries load, removal loses meaning
- Adequate: Some devices present, mostly decorative
- Opportunity: Nothing doing double duty
5. Human-Moment Anchoring
What: Abstractions grounded in specific scenes or lived experience. The concrete moment that earns the right to abstract.
Detection rules:
- Abstract argument without a grounding moment = opportunity
- Opens with specifics before generalizing = strong
- Physical language for abstract concepts = strong
- "Switching costs" vs the specific moment someone couldn't leave = difference
Rating:
- Strong: Abstractions earned through concrete moments
- Adequate: Some grounding, some floating
- Opportunity: Arguments proceed without lived anchoring
Output Format
Craft Review Table
Overall Assessment
One sentence: how memorable is this piece? Would a reader remember anything from it a week later?
Constraints
- Never rewrite prose yourself
- Proposed improvements are directional, not prescriptive
- "N/A" is valid for short pieces (< 200 words) on some dimensions
- Do not confuse literary ornamentation with structural devices
- Rate against the piece's own ambition (a quick email has different expectations than a blog post)
Before you install
- Read the whole file first. Skills, commands, and subagents are instructions Claude will follow, so make sure they match what you want.
- Check which tools, scripts, or MCP servers it uses. Local servers and scripts run with your permissions.
- Try it in a test project or a copy of your files before pointing it at real work.
- Pin the version you tested, and review changes before updating.
- Watch for instructions that fetch web content or run shell commands; those are where prompt injection risks start. See our prompt injection guide.
FAQ
What is Craft Reviewer?
Craft Reviewer is a subagent for Claude Code and Claude Cowork from the athola/claude-night-market repository on GitHub. Evaluate generated text for craft depth across five dimensions -
How do I install Craft Reviewer in Claude Code?
Download craft-reviewer.md from the repository. Save it to ~/.claude/agents/ to use it in every project, or to .claude/agents/ inside one project to share it through version control. Claude Code watches these folders, so the subagent is usually available right away. Ask Claude to use it by name, or @-mention it to make sure it runs.
Can I use Craft Reviewer in Claude Cowork?
Cowork loads subagents through plugins. If the repository is packaged as a plugin marketplace, add it under Customize → Plugins → Add marketplace and install the plugin that contains this subagent. Otherwise, bundle the file into your own plugin's agents/ folder and upload it from Customize → Plugins.
Is Craft Reviewer safe to install?
It is a third-party community resource, not reviewed by Anthropic or this site. Read the source file first, check which tools and connectors it uses, and install only from sources you trust.
Similar resources
- Code Refiner Orchestrate code quality refinement: analyze living code for duplication, algorithmic inefficiency, clean code violations, and architectural misfit. Generate concrete refactoring plans with before/after examples. Use PROACTIVELY for: code quality improvement, refactoring sprints, anti-slop remediation PRE-INVOCATION CHECK (parent must verify BEFORE calling): - "Fix this one function"? -> Parent… Subagent · athola/claude-night-market
- Catchup Context recovery after session breaks. Summarize recent git changes, extract key decisions, and identify pending work. Slash Command · athola/claude-night-market
- Code Review Mode Main thread configuration for evidence-based code review sessions. Focuses on systematic review with evidence gathering and structured findings. Use via: claude --agent code-review-mode Or set in .claude/settings.json: { "agent": "code-review-mode" } Subagent · athola/claude-night-market
- Code Searcher Search GitHub for existing implementations of a research topic. Returns structured findings with repo metadata, pattern analysis, and relevance ranking. Lightweight agent scoped to code search only. Subagent · athola/claude-night-market
- Dependency Updater Dependency analysis and update agent for multi-ecosystem repositories. Subagent · athola/claude-night-market
- Continuation Agent Lightweight agent designed to continue work from a session state checkpoint. Spawned when the parent agent exceeds context thresholds. This agent: 1. Reads the session state file 2. Re-establishes necessary context 3. Continues the task without interruption 4. Can spawn another continuation agent if needed Subagent · athola/claude-night-market
- Desktop Pilot Autonomous desktop control agent using Claude's Computer Use API. Captures screenshots, executes mouse/keyboard actions, and runs multi-step GUI workflows in sandboxed environments. Subagent · athola/claude-night-market
- Context Optimizer Autonomous agent for context window optimization and MECW compliance. Subagent · athola/claude-night-market