- Home
- Skills
- Workflow Automation
- Upgrade Loop Orchestrator
More screenshots
Works with the AI Agents you already use
Upgrade Loop Orchestrator
Run a scheduled loop that improves your codebase in small gated branches — queue, briefs, reports, digests — while merges stay human-approved.
$5
Upgrade Loop Orchestrator
Example session with this skill installed
Set up a continuous upgrade loop for my Rails side project. I can review and merge on weekends only, and I don't want an agent ever pushing to main.
- Read your context and instructions
- Compiled the upgrade loop orchestrator
Here is your loop, set up for weekend-only merging.
Directory: upgrade-loop/ in the repo, holding queue.json, state.json, phases/, and weekly digests.
Seed queue (3 phases, each with a full brief — deliverables, must-not-change, base, gates):
extract-invoice-pdf— move PDF generation out of the controller into a service object. Must not change: routes, mailer behavior. Gates:bin/rails testall passing (412 baseline), RuboCop clean on touched files.n-plus-one-sweep— fix the three N+1 queries flagged in the dashboard controller. Must not change: any rendered output. Gates: full suite + a request spec proving identical response bodies.dead-code-pass— remove the unused legacy reporting module behind its flag. Gates: full suite + production build.
Tick behavior: every 6 hours a tick reads state, verifies no builder is still alive (process + branch commits, not just the absence of a report), and spawns at most one builder in a fresh worktree. The builder runs the gates, commits to feature/upgrade-<slug>, and stages phases/<slug>/REPORT.md with the actual test counts.
Your weekend digest lists merge-ready branches in order, what each report proved, and anything disclosed — pre-existing failures or flaky tests, with evidence. Nothing merges or pushes until you say so; the loop pauses while you merge and resumes on the new main.
Connects securely to your tools. The creator never sees your data.
About this skill
Most codebases do not stall for lack of ideas; they stall because improvement work arrives as vague intentions and lands as risky, unreviewable changes. The Upgrade Loop Orchestrator turns that intent into a standing system: a queue of written briefs, one builder at a time on a fresh worktree and feature branch, the project's full gate suite before anything is called done, and periodic digests that show a human exactly what is merge-ready, in what order, and on what evidence.
The loop is deliberately conservative where it matters. It never merges and never pushes. Every phase ends as a branch plus a report carrying real gate numbers, disclosed pre-existing failures, and flaky checks recorded with both runs — so the human's merge decision takes minutes and rests on evidence, not summaries. The skill covers the full operating discipline: the state files the loop owns, the eight-step tick workflow, how to verify a builder is actually alive before re-queuing its phase, how to salvage interrupted work instead of restarting it, brief and digest templates, a tick checklist, an alive/dead decision table, and the failure modes that real loops hit — overlapping ticks, API-push ancestry that lies about what merged, and queues refilled with filler.
It fits any project with a runnable gate suite (typecheck, tests, build) and any scheduler — cron, a CI schedule, or manual ticks. If you run coding agents and want steady, reviewable progress between your own sessions without handing an agent the keys to main, this is the operating pattern.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Workflow Automation.
- The loop is only as good as the project's gate suite; a codebase with weak tests gets well-organized branches whose reports honestly say the gates could not prove much.
- It assumes one human (or team) actually gives merge words on a cadence — merge-ready branches that nobody merges become a stale backlog.
- It does not diagnose or repair a loop that is already failing in production; that is a separate loop-health audit job.
- Brief quality is the ceiling: vague briefs produce vague phases, and the skill says so rather than papering over it.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 24 days ago
Permissions
No special permissions declared or detected