# loop-converge

Adopt this skill: https://proskills.md/skills/other/levi-qiao-longgraph-skill-skills-loop-converge

Page: https://proskills.md/skills/other/levi-qiao-longgraph-skill-skills-loop-converge
This document is the public skill bundle for that page. It contains SKILL.md and every helper file the catalog carries. No login is required.

## File: skills/loop-converge/SKILL.md
size: 1843
upstream: https://github.com/levi-qiao/longgraph-skill/blob/main/skills/loop-converge/SKILL.md

```text
---
name: loop-converge
description: Use only when the user explicitly invokes /loop-converge or names this skill. Authors a loop-converge run. Do not use unless named. Do not execute or resume generated runtime node files.
disable-model-invocation: true
---

# loop-converge — a loop-graph preset for slimming code

A thin authoring entry. It binds a code-convergence pack, starts the owner
interview, then follows [`loop-graph`](../loop-graph/SKILL.md) to compile a
normal two-node run (executor + supervisor). It is not a third runtime node
and it does not ship a second template set.

## Fit check

- Use this when unused or duplicate code will take many verified rounds to
  remove, merge, or reuse, and an independent audit should keep the bar from
  sliding into a rewrite.
- If the request is a one-shot tidy that fits a normal host task, say so and
  send the task directly. Do not wrap it in a graph.

## On invoke

1. Inspect the workspace and the current host the same way loop-graph does.
   Never ask which client this is when context already identifies it.
2. Read and bind [`preset.md`](preset.md). That pack **is** the North Star,
   the supervisor requirement, the method guards, the knob overrides, the
   recommended shape, and the detector hints. Do not redesign them.
3. Start the owner interview immediately. With this pack bound, ask at most
   the three questions the pack names — **scope**, **authority**, **launch
   mode** — each as a recommended A/B (or A/B/C) choice. Do not ask for a
   North Star. Do not offer to omit the supervisor.
4. Read and follow [`../loop-graph/SKILL.md`](../loop-graph/SKILL.md) from
   **When called from a preset skill** through generate and deliver. Compile
   only from loop-graph's [`templates/`](../loop-graph/templates/). This skill
   never executes the generated nodes.
```

## File: README.md
upstream: https://github.com/levi-qiao/longgraph-skill/blob/main/README.md

```text
# loop-converge

**A thin `/loop-converge` entry that starts a code-convergence interview, then compiles a normal loop-graph run.**

English · [简体中文](README.zh-CN.md)

`/loop-converge` is not a second runtime. It binds a preset pack (unused /
duplicate / reuse / merge / slim; supervisor required) and follows
[loop-graph](../loop-graph/SKILL.md) to generate executor, supervisor, ledger,
and directives under `.longgraph/<date>-converge/`.

## When to use it

Multi-round cleanup where "done" is checkable: tests stay green, detectors
drop, net lines do not grow, no new public surface. The run sweeps the scope in
rotation and discovers its own candidates each round — it ends when fresh
detector passes stop yielding, not when a seeded list runs out. Skip it for a
one-shot tidy. One round lands the largest safe related workset sharing a proof
bar and gate; a fire cap only starts the next fire, never ends the run.

## How to run it

1. Invoke `/loop-converge` (Claude Code plugin: `longgraph:loop-converge`).
2. Answer the short interview — scope, authority, launch — or accept the
   recommended A on each.
3. **A** creates both nodes on Codex or Claude Code. **B** prints copy-ready
   read-and-follow prompts for another host, including Grok Build.

Authoring never executes the generated nodes.

## Files

| Path | What it is |
| --- | --- |
| [`SKILL.md`](SKILL.md) | Entry: fit check, bind the pack, start the interview, hand off |
| [`preset.md`](preset.md) | Authoring-only pack — North Star, guards, knobs, detector hints |

Runtime templates live only under [`../loop-graph/templates/`](../loop-graph/templates/).
```
