菜单
Reference for how to use the repowire mesh — ask/ack vs notify, broadcast, peer discovery, spawning, and cross-agent workflows. Use when you need to coordinate with other AI agents over repowire and want the right primitive for the job.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Parallel divergent ideation via mesh fan-out. Spawns N frame-shifted peers, scores results, deepens top survivors. Use on open-ended design choices, fuzzy bugs, naming, API surfaces, or explicit /adhd invocation. Skip for syntax, lookups, or closed prompts ("quick", "standard", "canonical").
Get an independent implementation plan from a DIFFERENT AI agent over the repowire mesh before you build (e.g. have Gemini or Codex draft an approach for Claude to critique). Use when you want a second perspective on how to approach a task.
Get a code/PR/plan review from a DIFFERENT AI agent over the repowire mesh (e.g. have Codex review Claude's work). Use when you want an independent second opinion from another backend before merging or committing.
Hand a task to another AI agent peer over the repowire mesh — reuse an existing peer on the chosen backend, or spawn one if none exists. Use when you want to offload work to a specific agent (e.g. delegate a refactor to a Codex peer) and track it to completion.
Install or update repowire and its skill pack from inside an agent session — ensure the repowire CLI/hooks are set up, install the skills, and verify the mesh works. Use on a fresh machine or to refresh skills to the latest.
Use when orchestrating complex or multi-lane work, decomposing tasks across peers, choosing status/board surfaces, running mesh roundups, or keeping high-level progress visible.
| name | repowire-patterns |
| description | Reference for how to use the repowire mesh — ask/ack vs notify, broadcast, peer discovery, spawning, and cross-agent workflows. Use when you need to coordinate with other AI agents over repowire and want the right primitive for the job. |
Repowire is a mesh where AI coding agents (Claude Code, Codex, Gemini, OpenCode,
Pi, …) each get an address and talk to each other. This is a teaching reference;
the action skills (cross-agent-review, cross-agent-plan, delegate) apply
these patterns. It does not depend on those skills being installed.
| Want | Use | Lifecycle |
|---|---|---|
| Tracked request that needs a reply | ask(peer, text) → peer ack(corr_id, reply) | Non-blocking; returns correlation_id; reopen with ask(reply_to=...) |
| Fire-and-forget nudge / status | notify_peer(peer, text) | No reply expected |
| Message everyone in your circle | broadcast(text) | No per-peer lifecycle |
| Close an inbound ask | ack(corr_id) bare, or ack(corr_id, reply) | The only close/reply op for an ask |
| See who's around | list_peers() | Returns names, backends, status |
| Who am I | whoami() | Your peer identity/circle |
The tracked ask/ack/notify/broadcast lifecycle is MCP-only — use the
mcp__repowire__* tools. There is no honest CLI equivalent for non-blocking ask
(the repowire peer ask command is a synchronous testing utility, not the
ask/ack lifecycle). For agents without MCP, the CLI offers these real fallbacks:
repowire peer list, repowire peer whoami, repowire peer ack <cid>,
repowire peer asks (list pending), repowire peer new (spawn).
ask is non-blocking — it returns a correlation_id, not a reply. The reply
arrives later as an ack. Don't wait synchronously.ack is the only way to close an inbound ask. Bare ack(corr_id) = "seen,
no action"; ack(corr_id, msg) = reply. Unacked asks resurface as reminders.peer_id; addressing is display_name. Names can collide; pass
circle to disambiguate.cross-agent-review/cross-agent-plan).spawn_peer starts a new agent; confirm with the user
rather than spawning silently.cross-agent-review (have a different backend review your work).cross-agent-plan.delegate (reuse or spawn a peer, hand off, track via ack).Backends for these are parameterised via repowire config get skills.* defaults,
overridable per call — never hardcode a backend.