Skip to content

Blog

Vouch for the harness, not for its output

Code review cannot keep up with agent output. Open source maintainers already admit contributors instead of reviewing contributions, and in an agentic factory the contributor to admit is the harness.

Written by

Agent security and risk

Published: 21 April 2026 Updated: 1 min read

Every team is trying to keep up with code review. The right move is to make it obsolete.

More reviewers will not do it. A factory that admits trusted contributors, and only trusted contributors, will.

Admit the contributor, not the contribution

A new open-source mechanism called Vouch lets maintainers explicitly vouch for each human contributor. No vouch, no PR. The decision moves from “review the contribution” to “admit the contributor.”

That is half the pattern.

In a factory, the contributor is the harness

In an agentic factory, the contributor is not a human. It is the harness: a specific, versioned combination of model, tools, prompts and quality gates that produces code.

Same weights, same training, different harness, different output.

If the harness is the contributor, the harness is what you vouch for.

What a vouched harness looks like

  • Named and versioned. Not “Claude with some MCP servers,” but a specific config at a specific version.
  • A track record. It shipped, reliably, under your quality bar.
  • Scoped. Trusted to ship feature flags, not to touch migrations.
  • Promoted, not upgraded. A new model version does not inherit the trust of the old harness. It earns its own.

If its output passes the gates, it ships, with no review. The vouch replaces the review because the contributor was approved before the keyboard touched the repo.

Why this holds

Above a certain complexity threshold, human code reviews decrease software quality. Reviewers miss things when PR volume explodes, and tests catch more than eyes do. A factory with tight quality gates and a vouched harness produces output that is already verified, at scale, without a human in the middle.

The agent stays small and the factory does the rest. The harness is the unit of trust.

Are you reviewing harness output forever, or vouching for the harness?

Further reading

  1. Every automation needs a scoped, reversible sandbox Teams design what their automations do and neglect where they run. A written blast radius, a narrow scope and throwaway environments decide whether you can leave an automation unattended. 2 min readShort post
  2. A vibecoded SaaS leaked user data because nobody owned the design A production SaaS built by vibe coding failed in four different ways with one root cause. What survived shows why AI-generated code needs someone who owns the system design. 1 min readShort post
  3. Well-built agents close compliance gaps that manual processes leave Compliance teams block agents as too risky while manual processes run on memory and reconstructed logs. Properly architected agents make audit trails and enforcement automatic. 2 min readShort post

All 6 posts on Agent security and risk

ClientsClient work

Scaling needs named owners

Coding agents scale across teams when named people own the platform, the team practice, the guardrails, and the measurement. Rollout & scale is where that ownership is built and handed to internal owners, in your stack.

Training on this topic: Agentic Engineering Security for Development Teams
See Rollout & scale

Want to check one team first? The self-check: twelve yes or no questions, three minutes, instant result, no account. Take the self-check