om-auto-continue-pr-loop

Resumes unfinished loop-created pull requests from the first incomplete PLAN.md task.

1|Updated Apr 23, 2026
One-click install
npx skills add https://github.com/tkogut/agents-os-core --skill om-auto-continue-pr-loop-tkogut
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: om-auto-continue-pr-loop
Source: https://github.com/tkogut/agents-os-core/tree/main/global_skills/om-auto-continue-pr-loop
Command: npx skills add https://github.com/tkogut/agents-os-core --skill om-auto-continue-pr-loop-tkogut

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Long-running, multi-step pull requests created by om-auto-create-pr-loop often cannot finish in a single session, leaving teams with half-done branches, unclear progress state, and no safe way to pick the work back up without duplicating effort or clobbering another agent's lock. ## Core Features & Use Cases - Stateful resume from PLAN.md: Claims the PR with a three-signal lock, reads HANDOFF.md and the Tasks table, and resumes from the first row whose status is not done. - Lean per-Step commits with checkpoint verification: Lands one commit per Step, batches validation, integration tests, and UI screenshots into checkpoint passes every 5 Steps, and runs a full final gate before marking the PR complete. - Lock discipline and chained hand-offs: Transfers locks between chained skills without releasing them mid-run, hands spec-only PRs to om-auto-implement-spec, and always releases the lock even on failure. - Use Case: A 12-Step feature PR stalled at Step 7 overnight; invoke this Skill with the PR number and it re-enters the isolated worktree, completes Steps 7-12 with checkpoints, runs the review pass, and flips the draft PR to ready. ## Quick Start Resume the unfinished loop run on pull request 1492 by running om-auto-continue-pr-loop with that PR number.

Frequently Asked Questions about om-auto-continue-pr-loop

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I resume an unfinished automated pull request?▼

Pass the PR number to om-auto-continue-pr-loop. It claims the PR, reads HANDOFF.md and the PLAN.md Tasks table, and resumes from the first row whose status is not done, driving the PR to completion with per-Step commits and checkpoint verification.

What is the difference between om-auto-continue-pr and om-auto-continue-pr-loop?▼

om-auto-continue-pr-loop resumes PRs created by om-auto-create-pr-loop that carry a run folder with PLAN.md, HANDOFF.md, and NOTIFY.md. Use plain om-auto-continue-pr for simple runs or PRs with no resolvable plan; it reconstructs a plan from PR context.

What happens if another agent already claimed the pull request?▼

The skill runs a three-signal in-progress check covering assignee, in-progress label, and recent claim comments. If another actor holds the lock it stops and asks, unless you pass --force, which posts an override comment naming the previous owner before taking over.

Does the skill release the lock if the run crashes?▼

Yes. Lock release is wrapped in a trap/finally so the in-progress label is removed and a completion comment is posted even on failure. Stale locks older than the defined windows can be recovered with a takeover note.

When does a resumed pull request get marked ready for review?▼

The PR stays a draft while its status is in-progress. Only after every Tasks row is done, the full validation gate and integration suite pass, and the om-auto-review-pr autofix pass is clean does the skill flip the PR from draft to ready.

Can a spec-only design PR get implementation code during a resume?▼

No. A spec-only guard stops the resume when remaining tasks would land implementation code on a design-only PR. The skill hands off to om-auto-implement-spec, which ships the implementation on its own PR referencing the spec PR.