Agent trajectory
Messages, tool calls, actions, artifacts, and the path taken to complete the task.
Reading note · Agent systems
LangChain’s Eval Engineering Skill turns a repository and its agent traces into proposed, executable evals. The useful part is that the process still makes room for the builder to decide what is actually worth measuring.
Read Viv’s original article on X ↗The skill inspects the agent surface, uses traces as observed contracts, proposes abilities, and iterates with the user before producing the final task.
Read prompts, models, tools, skills, hooks, data, and services.
Observe real arguments, results, errors, requests, and failures.
Choose the abilities that matter and decide what runs live or simulated.
Create the task, inspect the result, revise the verifier, and compare again.
evals/<task-id>/ ├── task.toml ├── instruction.md ├── environment/ └── tests/
The task the agent is asked to complete.
A controlled setup with the tools, data, permissions, and state the task needs.
The scoring logic that decides whether the intended capability was demonstrated.
A first passing score is not enough. The article recommends inspecting both sides of the result to see whether the task measured the thing the builder actually cared about.
Messages, tool calls, actions, artifacts, and the path taken to complete the task.
Evidence, reasoning, scoring, and the assumptions that turned behavior into a reward.
The strongest part of this approach is that it does not pretend the agent can discover the right measurement in one pass. Repository structure and traces provide evidence, but the builder still has to say which capability matters and whether the environment represents the real work.
An eval becomes useful when it captures a real capability in a reproducible environment, not when it simply produces a clean score.
I keep coming back to the maintenance question. I want failures and unexpected behavior to become durable tests, but I do not want another layer of configuration that grows faster than the work it is protecting. The interesting version is a small loop where real traces identify the few evals that have earned a permanent place.
Which recurring failures or abilities deserve a durable eval instead of ordinary review feedback?
Does the verifier reward the intended result, or can the agent satisfy a proxy and still receive full credit?
Can the environment and verifier stay representative without becoming another large system to maintain?