cool-down

Stabilizes overloaded development batches by settling in-flight branches, agents, and tracking documents.

1|Updated May 6, 2026
One-click install
npx skills add https://github.com/diy-accounting-uk/submit.diyaccounting.co.uk --skill cool-down-diy-accounting-uk
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: cool-down
Source: https://github.com/diy-accounting-uk/submit.diyaccounting.co.uk/tree/main/.claude/skills/cool-down
Command: npx skills add https://github.com/diy-accounting-uk/submit.diyaccounting.co.uk --skill cool-down-diy-accounting-uk

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When a development batch adds work faster than it lands it—multiple agents in flight, failing workflows, and a board that no longer reflects reality—this Skill provides a structured mode to stop new work, settle what is in flight, and later resume cleanly without losing anything. ## Core Features & Use Cases - Controlled cool-down mode: Freezes new board rows, parks discoveries in PARKED.md, and routes live customer-facing degradations to their own hotfix branch off main. - Branch-by-branch settling: Drives one branch green at a time, requiring local reproduction and proof of every fix before pushing, and checks for unpushed commits before calling a branch settled. - Document reconciliation and wake-up: Brings NEXT.md, BACKLOG.md, and plan documents back in step with reality, then defines an ordered eight-step wake-up sequence so resuming work never drifts from how it was paused. - Use Case: A team running parallel coding agents sees CI failing across several branches while new tasks keep arriving. Invoke cool-down to commit and stop agents, fix branches serially with locally proven fixes, update the tracking documents, and later run the wake sequence to resume dispatch safely. ## Quick Start Tell the agent to start cool-down mode for this repository and settle the in-flight branches and tracking documents.

Frequently Asked Questions about cool-down

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

FAQPage Schema
How do I stop a development batch that is failing faster than it lands?▼

Invoke cool-down mode, which writes a marker into NEXT.md, stops new board rows, tells running agents to commit and stop, and drives one branch green at a time. Discoveries are parked in PARKED.md for later triage instead of being lost or forced onto the board.

How do I handle a live customer-facing bug during a code freeze?▼

A live customer-facing degradation gets its own branch off main carrying only that fix, with its own PR. The operator is told to merge that PR first, before the batch, so the hotfix ships without waiting for anything else to be right.

Why should CI fixes be reproduced locally before pushing?▼

Pushing an unproven fix is a guess that costs a full CI cycle and turns a batch into a failure loop. Reproducing each failure locally and getting a passing result first is what stops the loop, so there is no cap on pushes once fixes are proven.

What happens to uncommitted work in git worktrees during cool-down?▼

Agents commit their current piece and stop, and each worktree is checked with git status --short. On wake-up, every remaining worktree is walked again so uncommitted work is recovered before any worktree removal destroys it.

When does cool-down mode end and normal work resume?▼

Only the operator lifts cool-down in their own words, such as carry on or resume. An eight-step wake sequence then runs: merge ready PRs, remove the marker, account for hotfixes and worktrees, confirm branches are green, read the documents, triage PARKED.md, and re-render the board.