AgentClash

Glossary

What is a release gate?

A release gate is the pass or fail boundary between a candidate agent run and an approved baseline. AgentClash evaluates both on the same pack and fails CI when correctness, cost, or evidence quality regresses.

live race
gate: pass

Candidate

92correct patch, low cost

Baseline

88stable reference run

Control

73missed edge case

replay timeline

1loaded task inputs and tool policy
2ran sandbox actions and captured artifacts
3scored trajectory and validator evidence
4attached scorecard and release verdict

ci verdict

Candidate clears release gate

Correctness improved, latency within budget, and required artifacts were preserved for review.

agentclash run create --follow

What gates enforce

Built for reviewable agent decisions

Scorecard thresholds, validator failures, and baseline comparisons that stop a model or harness change from reaching users.

Sandboxed real-tool execution

Head-to-head runs with fair constraints

Scorecards for correctness, cost, latency, and tool strategy

Replay trails for every important action

Challenge packs that turn failures into reusable tests

CI gates for baseline versus candidate decisions

Workflow

Gate workflow

Package the task

Describe the workload as a challenge pack with inputs, tools, scoring rules, and artifacts.

Race the agents

Run every candidate against the same task with the same constraints.

Replay the evidence

Inspect tool calls, outputs, artifacts, latency, cost, and judge evidence after the run.

Gate the release

Compare candidate and baseline runs, then fail CI before a regression reaches users.

FAQ

Release gate FAQ

What is a baseline in a release gate?

A baseline is the approved run you compare against, usually pinned by run ID in a CI manifest after a green eval on main.

Can gates run outside CI?

Yes. Teams also gate manually before production promotion, using the same scorecard and replay evidence produced in CI.

How is this different from a linter?

Linters check static code. Release gates execute the agent on real tasks and score the full trajectory with replay attached.