Skip to content

hariharapanigrahy/layerkit / skills/layerkit-multi-agent

layerkit-multi-agent

Verified

Coordinate specialist agents for integrate or contract heal; brownfield production code.

FileKindSize
SKILL.mdskill md2.2 KB
---
name: layerkit-multi-agent
description: Coordinate specialist agents for integrate or contract heal; brownfield production code.
---

# layerkit-multi-agent

Team of agents on the **same** integration pipeline. The lead agent coordinates specialists directly; Layerkit does not generate a deterministic multi-agent plan.

## Start

### Contract heal (map already exists — human supplies OpenAPI/docs)

- Discover task **omitted**
- Researcher = evidence-backed semantic rename decisions when needed
- If only docs are supplied, researcher/AI must read/cite docs before source edits
- AI agents edit production source/test files directly
- Record rename decisions in the handoff note only when docs/code evidence supports a removed-to-added field rename

### First-time / multi-vendor

Use the lead agent to assign vendors/files explicitly. Keep one writer per production file and use handoff notes for state.

## Roles

| Role | Skill | Heal note |
|------|-------|-----------|
| orchestrator | orchestrate-integration | status shows mode=heal\|full |
| discoverer | discover-data-layer | **not in heal plan** |
| stylist | align-client-style | topology for existing adapters |
| researcher | research-vendor | contract pin + drift |
| designer / author | design / processor | surgical on drift |
| privacy | privacy-review | human if new PII |
| integrator | source-edit-client | direct production source edits |
| verifier | doctor + project tests | quality |
| checker | checker-assist | read-only |

## Concurrency

- Parallel research/integrate per vendor when files do not overlap
- Single writer: `integrator:registry`
- Human: privacy, checker approval, breaking drift

## Forbidden

- Two agents on the same registry file
- Re-running full discover on heal when domain is known
- Invented fields or guessed renames; self-approve
- Treating generated plans as production source code
- Relying on Layerkit runtime/route/strategy code instead of editing the client package

## Success

- [ ] Agent assignments match intent (heal vs full)
- [ ] Specialists have explicit files/responsibilities
- [ ] Production adapters updated
- [ ] Project tests + quality green
- [ ] Human checker handoff complete
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall