GitHub-Native Workflows

End-to-end GitHub-native workflows -- from issue creation to merged PR.

Locus uses GitHub as an operational layer, not just a code host. Issues, Milestones, Labels, and Pull Requests carry planning and execution state that your whole team can inspect.


Workflow Overview

spinner

Scenario 1: Issue-Driven Delivery

The core delivery loop when you already know the task.

# Create sprint and issue
locus sprint create "Sprint 1"
locus issue create "Add /health endpoint with tests" --sprint "Sprint 1"

# Execute (auto-detects open sprints)
locus run

# Review and iterate
locus review
locus iterate --sprint

# Check progress
locus status

GitHub state after execution:

spinner
  • Milestone Sprint 1 tracks completion percentage

  • Issue labels update: locus:queuedlocus:in-progresslocus:done

  • PR created with Closes #83 so merging auto-closes the issue


Scenario 2: AI-Powered Planning

Let AI break down a high-level goal before execution.

What happens:

spinner
  • AI reads LOCUS.md, file tree, and recent git history

  • Generates ordered issues with priority, type, and execution sequence

  • Issues are created on GitHub with labels: p:*, type:*, locus:queued, agent:managed, order:N


Scenario 3: Review and Iteration Loop

When PR feedback needs to be addressed by AI.

What happens:

spinner
  • Review feedback is persisted as GitHub PR comments

  • Iteration updates PR branches with targeted fixes

  • Label/status lifecycle stays visible for async team coordination


Troubleshooting

gh is not installed or not authenticated

Not in a git repository (or missing GitHub remote)

locus run says no open sprints found

Rate limit or transient API failures

Wait for GitHub rate limit reset window, then rerun. For interrupted sprints:

Last updated