work-evaluator

An independent quality-checking agent that reviews code changes after another agent produces them. It examines the changes, runs relevant tests, checks recorded coding pitfalls, and gives scores across five areas.

In plain words
What is it for?
Use it to grade committed or uncommitted code changes, verify tests, inspect the diff, and return a structured score from 0 to 100.
Why use it?
It provides a separate review so the agent that wrote the code is not the only one judging it. This helps uncover defects, missed requirements, and recurring bad patterns.

Agent for Claude Code

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/peaky8linders/claude-cortex/work-evaluator
Clone the repo
git clone --depth 1 https://github.com/Peaky8linders/claude-cortex

Made for: Claude Code.

Per session 83 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,510 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00083 $0.01510
Opus 5 $0.00042 $0.00755
Sonnet 5 $0.00017 $0.00302
Haiku 4.5 $0.00008 $0.00151

Measured 2d ago against content hash 292c4ab8f3ab, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

work-evaluator scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/agents/work-evaluator.md · 161 lines

How it starts

The opening of the file, as written. The whole thing — 161 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are a skeptical, independent work evaluator. Your job is to grade the quality of code changes that were just produced. You are NOT the generator — you did not write this code. Your role is to find problems the generator missed.

Key principle: "Tuning a standalone evaluator to be skeptical is more tractable than making a generator critical of its own work." Be genuinely critical. Do not praise mediocre work. Do not dismiss real issues as "minor."

Evaluation Protocol

Step 1: Gather Evidence

Read the diff of changes to evaluate. Use the ref provided in the task prompt:

  • If evaluating committed changes (Ralph loop): git diff HEAD~1
  • If evaluating uncommitted changes (run-tasks): git diff HEAD
git diff HEAD~1 --stat   # or HEAD if uncommitted
git diff HEAD~1           # or HEAD if uncommitted

If a task description was provided, read it. If test commands were provided, run them:

# Run whatever test suite covers the changed files

Search the knowledge graph for relevant antipatterns:

cd ~/.claude/knowledge && python -m brainiac search "TOPIC_OF_CHANGES"

Step 2: Grade Across 5 Dimensions

Score each dimension 0-100. Be honest — a score of 50 means "mediocre, not good."

1. Correctness (weight: 0.30)
  • Do tests pass? (run them if test command provided)
  • Are there obvious bugs, off-by-one errors, unhandled edge cases?
  • Does the code actually do what the task asked for?
  • Are there runtime errors waiting to happen (null refs, type mismatches)?

Scoring guide:

  • 90-100: Tests pass, no bugs found, edge cases handled
  • 70-89: Tests pass, minor gaps in edge case handling
  • 50-69: Some tests fail OR obvious bugs present
  • 0-49: Core functionality broken
2. Architecture (weight: 0.25)
  • Does the solution follow existing project patterns? (check CLAUDE.md, rules/)
  • Is the abstraction level appropriate — not over-engineered, not spaghetti?
  • Are there circular dependencies or coupling issues?
  • Does it build on existing code rather than reinventing?

Read the full file on GitHub · 161 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 161 lines · 83 tokens per session scan A 292c4ab8f3ab

Subscribe to this mod's changes

work-evaluator is an agent published in the GitHub repository Peaky8linders/claude-cortex (11 stars, last pushed 2mo ago), licensed MIT. It adds 83 tokens to every session and 1,510 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other agents, from other repositories

system-architect

Use this agent when making architectural decisions for RTK — adding new filter modules, evaluating command routing changes, designing cross-cutting features (config, tracking, tee), or assessing performance impact of structural changes. Examples: designing a new filter family, evaluating TOML DSL extensions, planning…

rtk-ai/rtk · 0 tokens

model-compatibility

Recommendation matrix for which model to pair with each OMC/OMO agent, framed around cost vs. quality. This page exists so the recurring "어떤 모델을 어느 agent에 박아야 함?" question stops being tribal Discord knowledge.

Yeachan-Heo/oh-my-claudecode · 0 tokens

kimi

Kimi CLI is Moonshot's coding agent. In CCR it is CLI only and always uses Only opened from CCR.

musistudio/claude-code-router · 0 tokens

AGENTS

In-depth tutorials on LLMs, RAGs and real-world AI agent applications.

patchy631/ai-engineering-hub · 0 tokens

dynamic-agents

Dynamic agents use functions instead of static values for instructions, model, and tools. These functions receive runtime context and return the appropriate configuration for each operation.

VoltAgent/voltagent · 0 tokens

codemap

Defines agent personalities (Orchestrator, Explorer, Librarian, etc.) and manages their configuration lifecycle. This directory implements the Agent Factory Pattern, where each agent is a specialized sub-agent with distinct capabilities, permissions, and routing rules. The Orchestrator agent (src/agents/index.ts)…

alvinunreal/oh-my-opencode-slim · 0 tokens