evalyn-eval

A workflow for evaluating an LLM agent project with evalyn, using recorded traces of the agent’s actions and results. It builds a test dataset, recommends measurements, and runs evaluations.

In plain words
What is it for?
Use it to build datasets from traces, inspect agent behavior, choose metrics for patterns such as tool calls or citations, and run evaluations.
Why use it?
It replaces abstract guesses about quality with tests based on what the agent actually did. It also checks whether the project has recorded traces before evaluation begins.

Skill for Claude CodeCodex

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 skills/shihongdev/evalyn/evalyn-eval
Any agent
npx skills add shihongDev/evalyn --skill evalyn-eval
Clone the repo
git clone --depth 1 https://github.com/shihongDev/evalyn

Made for: Claude Code, Codex.

Per session 26 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 792 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 $0.00026 $0.00792
Opus 5 $0.00013 $0.00396
Sonnet 5 $0.00005 $0.00158
Haiku 4.5 $0.00003 $0.00079

Measured 2d ago against content hash d2cc0731ea38, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

evalyn-eval 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.

sdk/skills/evalyn-eval/SKILL.md · 110 lines

How it starts

The opening of the file, as written. The whole thing — 110 lines — stays where its author put it; the contents beside it link to each section on GitHub.

evalyn-eval

Overview

Build a dataset from traces, auto-recommend metrics based on trace analysis, and run evaluation. This skill reads actual trace data to make metric recommendations rather than asking abstract questions.

Pre-flight

  1. Verify traces exist:
evalyn list-calls --limit 5

If no traces: "You need to instrument your agent first. Invoke evalyn-setup."

  1. Check if a dataset already exists:
ls data/*/dataset.jsonl 2>/dev/null

If dataset exists, skip to Step 2.

Step 1: Build Dataset

Identify the project name from the evalyn list-calls output (project column).

evalyn build-dataset --project <project-name>

Capture the output path - it prints "Wrote N items to ". Use this path for all subsequent commands.

Step 2: Auto-Recommend Metrics

Inspect a trace to understand the agent's behavior:

evalyn show-trace --last -v

Analyze the trace structure and recommend a bundle. Evalyn has 17 curated metric bundles:

Trace Pattern Recommended Bundle
Multiple tool calls, planning steps orchestrator
Tool calls + multi-turn context multi-step-agent
URLs or citations in output research-agent
RAG retrieval spans, source docs rag-qa
Conversational, multi-turn chatbot
Code blocks in output code-assistant
Short summary outputs summarization
Educational/tutorial content tutor
Content generation, blog posts content-writer
Customer-facing Q&A customer-support

To see all available bundles:

evalyn suggest-metrics --mode bundle --help

Apply the recommended bundle:

evalyn suggest-metrics --dataset <path> --mode bundle --bundle <recommended>

Then expand coverage with LLM-based selection from the full 130+ metric registry:

evalyn suggest-metrics --dataset <path> --mode llm-registry --append

This two-pass approach gives a solid base (curated bundle) plus tailored additions (LLM picks from full registry).

Read the full file on GitHub · 110 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 · 110 lines · 26 tokens per session scan A d2cc0731ea38

Subscribe to this mod's changes

evalyn-eval is a skill published in the GitHub repository shihongDev/evalyn (257 stars, last pushed 2mo ago), licensed MIT. It adds 26 tokens to every session and 792 once invoked, about $0.0001 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 skills, from other repositories

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

obra/superpowers · 21 tokens

next-cache-components-adoption

Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…

vercel/next.js · 95 tokens

babysit-pr

Babysit a GitHub pull request after creation by continuously polling review comments, CI checks/workflow runs, and mergeability state until the PR is merged/closed or user help is required. Diagnose failures, retry likely flaky failures up to 3 times, auto-fix/push branch-related issues when appropriate, and keep…

openai/codex · 114 tokens

imagegen

Generate or edit raster images when the task benefits from AI-created bitmap visuals such as photos, illustrations, textures, sprites, mockups, or transparent-background cutouts. Use when Codex should create a brand-new image, transform an existing image, or derive visual variants from references, and the output…

openai/codex · 113 tokens

cpu-profile-analysis

Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…

microsoft/vscode · 71 tokens

next-cache-components-optimizer

Drive a Next.js route to instant navigation by setting up an agentic loop, under Cache Components / PPR, on initial load (hard navigation) and client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route at a time; the shipped test then…

vercel/next.js · 170 tokens