Gates

Gates are checks that run after a step completes. They can be deterministic (your project’s build tools) or non-deterministic (a workflow validator backed by an agent).

Declaring gates

- name: impl
  type: code
  run:
    agent: claude-sonnet
  gate: [compile, test, lint]

Gates execute in order. All gates are evaluated even if one fails — the complete error context is available to the agent on retry. This lets the agent see all problems at once.

Built-in gates

Shell aliases

These resolve automatically based on your project type, or you can override them in gump.toml.

Gate	Go	Node	Rust
`compile`	`go build ./...`	`npm run build`	`cargo build`
`test`	`go test ./...`	`npm test`	`cargo test`
`lint`	`golangci-lint run`	`npm run lint`	`cargo clippy`
`coverage: N`	`go test -cover`	`npm test -- --coverage`	`cargo tarpaulin`

Resolution cascade: heuristic (project markers) → gump.toml → inline bash: gate. compile and test are required — they fail if they can’t be resolved. lint and coverage are optional — skipped with a warning if the tool isn’t available.

Structural gates

Gate	What it checks
`schema`	The split output is valid JSON: array of tasks with `name` and `description`
`touched: "glob"`	At least one file matching the glob was modified in the diff
`untouched: "glob"`	No file matching the glob was modified in the diff
`tests_found`	The test runner finds and recognizes tests to execute
`coverage: N`	Test coverage meets or exceeds N% (optional — skipped if tool absent)

Globs match on the basename, not the full path. "*_test.*" matches pkg/auth/middleware_test.go.

Custom gates

gate:
  - bash: "make integration-test"

Runs any shell command. Exit 0 = pass, non-zero = fail. Stderr is captured as error context for retries.

Workflow validators

A workflow with type: validate (output = bool) can be used as a gate. This is how agent-backed reviews work in v0.0.4:

gate:
  - compile
  - test
  - validate: validators/arch-review
      diff: "{diff}"
      spec: "{spec}"
      agent: claude-opus

The validator workflow executes its own GET → RUN → GATE cycle. Its comments are accessible via {gate.review.comments} on retry. See Workflow Composition for details.

Combining gates

Gates are a list. Use as many as you need:

gate:
  - compile
  - test
  - "untouched: *_test.*"
  - "coverage: 80"
  - bash: "make smoke"
  - validate: validators/arch-review
      diff: "{diff}"
      agent: claude-opus

Overriding detection

If auto-detection picks the wrong commands, override in gump.toml:

[validation]
compile_cmd = "make build-affected"
test_cmd = "make test-unit"
lint_cmd = "golangci-lint run --timeout 5m"

Config overrides apply to all workflows in the project.

Overview

Getting Started

Writing Workflows

Orchestration

Execution & Data

Agents

Reference

Gates

Gates

Declaring gates

Built-in gates

Shell aliases

Structural gates

Custom gates

Workflow validators

Combining gates

Overriding detection

​Gates

​Declaring gates

​Built-in gates

​Shell aliases

​Structural gates

​Custom gates

​Workflow validators

​Combining gates

​Overriding detection

Gates

Declaring gates

Built-in gates

Shell aliases

Structural gates

Custom gates

Workflow validators

Combining gates

Overriding detection