DEV Community

Cover image for wf-verify: One Verify Gate for Every Ecosystem
Faris Han
Faris Han

Posted on Edited on

wf-verify: One Verify Gate for Every Ecosystem

wf-verify: One Verify Gate for Every Ecosystem

Agentic loops break down at the verify step. The agent writes code, then either
skips verification or runs a command that only covers one ecosystem. A Go repo
gets go test; a Bun repo gets bun test; a mixed repo gets nothing consistent.
The result: "it passed" means different things in different projects.

wf-verify is one command that detects the project and runs the right gate.

What it does

  • Detects project type from files present (go.mod, package.json, Cargo.toml, pyproject.toml, Makefile, and others).
  • Runs the appropriate build, lint, and test commands for each detected ecosystem.
  • Returns a single pass/fail exit code the loop can gate on.
  • Works in CI and locally with the same invocation.

Coverage: Go, Bun/Node, Rust, Python, Make, and more (9 ecosystems total).

Why a single command matters

In an agentic loop, the verify step is the contract between "implemented" and
"shipped." If the command differs per repo, agents (and humans) improvise, and
improvisation is where regressions slip in. One command means:

  • The agent cannot quietly skip a step.
  • CI and local agree.
  • Reviewers know exactly what "verified" means.

Using it in the loop

# after implementing
wf-verify
# exit 0 -> proceed to review
# exit non-zero -> fix, do not review or ship
Enter fullscreen mode Exit fullscreen mode

Make it mandatory in your agent commands: /verify runs wf-verify and blocks
/ship on failure. That single rule removes most "shipped broken code" incidents.

Install

wf-verify ships with the OpenCode Dev Toolkit (Pro). The core workflow kit
(MIT) is free on GitHub.

Takeaway

Verification is only a gate if it is the same gate every time. One command,
every ecosystem, same result.

Top comments (0)