Harness Engineering

handfuloflight1 pts0 comments

GitHub - lopopolo/harness-engineering: Ryan Lopopolo’s anthology, field guide, and agent context bundle for harness engineering · GitHub

/" data-turbo-transient="true" />

Skip to content

Search or jump to...

Search code, repositories, users, issues, pull requests...

-->

Search

Clear

Search syntax tips

Provide feedback

--><br>We read every piece of feedback, and take your input very seriously.

Include my email address so I can be contacted

Cancel

Submit feedback

Saved searches

Use saved searches to filter your results more quickly

-->

Name

Query

To see all available qualifiers, see our documentation.

Cancel

Create saved search

Sign in

/;ref_cta:Sign up;ref_loc:header logged out"}"<br>Sign up

Appearance settings

Resetting focus

You signed in with another tab or window. Reload to refresh your session.<br>You signed out in another tab or window. Reload to refresh your session.<br>You switched accounts on another tab or window. Reload to refresh your session.

Dismiss alert

{{ message }}

lopopolo

harness-engineering

Public

Notifications<br>You must be signed in to change notification settings

Fork

Star<br>36

trunk

BranchesTags

Go to file

CodeOpen more actions menu

Folders and files<br>NameNameLast commit message<br>Last commit date<br>Latest commit

History<br>2 Commits<br>2 Commits

assets

assets

docs

docs

evals

evals

playbooks

playbooks

sources

sources

.gitattributes

.gitattributes

.gitignore

.gitignore

.prettierignore

.prettierignore

.prettierrc.json

.prettierrc.json

AGENTS.md

AGENTS.md

ARCHITECTURE.md

ARCHITECTURE.md

CLAUDE.md

CLAUDE.md

COPYING.md

COPYING.md

LICENSE

LICENSE

README.md

README.md

View all files

Repository files navigation

Harness Engineering

“Most people do not know that they can just point their agents at my writing,<br>tweets, podcasts, and talks and improve the output of their agents by 100x.”

— Ryan Lopopolo

Harness engineering, the practice of improving agent output by shaping the<br>environment around it, holds a chosen model and coding agent constant as a black<br>box. It improves the two external levers—context and tools—and curates the<br>environment around them. The worker should be able to recover intent, operate<br>the real system, respect authority, prove the outcome, and leave the next run<br>better equipped.

A central purpose of that environment is to carry an organization's<br>nonfunctional requirements: the quality attributes and constraints governing<br>reliability, security, compatibility, maintainability, performance, operability,<br>risk posture, and polish. The harness also carries local decisions about how to<br>prioritize, trade off, and satisfy those requirements. Ryan adopted a<br>systems-level framing from 2026’s [un]prompted conference that describes this<br>as getting the whole universe of nonfunctional requirements into code. Make the<br>Repository Teach the Agent develops how the requirements and decisions become<br>retrievable context, examples, tools, and executable constraints.

Because work is an iterative game, a harness can make organizational judgment<br>cumulative. Lessons from accepted work, corrections, failures, and user<br>responses become context, boundaries, tools, examples, and checks that shape<br>later trajectories. Over time, that feedback loop can make coherence<br>cumulative across agent-maintained artifacts.

Code is how an agent uses a computer. That internal action language can<br>produce reliable domain outcomes for people who never review the implementation<br>when last-mile deployment supplies the organization’s context, capabilities,<br>authority, and proof.

General model weights contain only the visible tip of an organization’s<br>process-data iceberg. Below the waterline sit the current operational state,<br>local ontology, quality bar, procedures, exception history, and authority<br>relationships that an agent needs to do a particular job. Organizations cannot<br>presume that this private, changing process data will be present in general<br>model weights, nor that agents will reliably intuit which process data matters<br>to the organization. Harness engineering is the last-mile work of making it<br>available to a capable worker as context and tools.

Point a coding agent at this repository alongside the system it should improve.<br>AGENTS.md routes the task to the relevant arguments, cases, and proof. For<br>direct reading, start with the thesis index. For an application, choose from<br>the playbooks.

Sources and related work

“Harness engineering: leveraging Codex in an agent-first world” (fetch<br>helper for agents blocked by the canonical page)

Source library

Influences and alternate framings

Repository-authored material is licensed under CC BY 4.0. See COPYING.md<br>for attribution and rights in source material.

About

Ryan Lopopolo’s anthology, field guide, and agent context bundle for harness engineering

openai.com/index/harness-engineering/

Resources

Readme

License

CC-BY-4.0, Unknown licenses found

Licenses found

CC-BY-4.0

LICENSE

Unknown

COPYING.md

Uh oh!

There...

harness engineering agent context agents lopopolo

Related Articles