multi-agent-quickstart
Use when learning how to reach Codex from this Windows-first setup - the beginner mental model for the codex mcp-server rail (mcp__codex__codex / mcp__codex__codex-reply), how the retired per-mode codex_* tools map onto prompt framings, when to fan out parallel calls instead, and how to choose an execution path after planning.
- License 许可
- MIT
- 最近更新
- 2026-11-02
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- multi-agent-quickstart
- description
- Use when learning how to reach Codex from this Windows-first setup - the beginner mental model for the codex mcp-server rail (mcp__codex__codex / mcp__codex__codex-reply), how the retired per-mode codex_* tools map onto prompt framings, when to fan out parallel calls instead, and how to choose an execution path after planning.
02 资源文件 1
- 参考文档agents/openai.yaml179 B
03 SKILL.md 正文
Multi-Agent Quickstart
Rewritten 2026-08-09. The previous version taught six MCP tools (codex_consult,
codex_brainstorm, codex_review, codex_discuss, codex_coordinate, codex_curate) plus
an optional PromptBridge talk lane. All of them came from tools/mcp/codex-orchestrator/ and
tools/web/promptbridge-talk-lane/, both deleted. These are not renames - the old tool names
fail rather than degrade.
Mental model
There is one rail: the official codex mcp-server.
mcp__codex__codex- start a Codex thread.mcp__codex__codex-reply- continue bythreadId(hyphenated; the underscore form does not resolve).
The old tool names are now prompt framings on that one rail:
| Old tool | Now |
|---|---|
codex_consult |
one call asking for one strong answer |
codex_brainstorm |
one call asking for N alternatives |
codex_review |
one call handing over the plan/diff, asking for a critic pass |
codex_discuss |
two parallel calls with opposed framings, synthesized by you |
codex_coordinate |
one call per role in parallel, consolidated by you |
codex_curate |
one call handed the option set, asked to select and justify |
Fan-out replaces the orchestration server. Each call is single-threaded and fresh-context, so anything that must persist belongs in the prompt or in a file.
Beginner flow
- Make one
mcp__codex__codexcall and read the answer. - Ask for alternatives in a second call if one answer is not enough.
- Ask for a critic pass before a risky implementation.
- Hand the option set back for selection when choosing among options.
Execution choice
- After planning, prefer subagent-driven execution for complex work.
- Use inline execution when the task is small or subagents would be overhead.
- Do not silently force one path after the plan is written.
Failure behavior
There is no second lane to fall back to, and no fallback_to_lane_a flag. If the rail is
unavailable, say so and stop - do not answer as though a Codex peer had been consulted.
Diagnose with codex --version and claude mcp get codex (expect Connected).
Related skills
codex-orchestrate (mode routing on this rail) · codex-invoke (exec rail, per-call
scoping) · codex-consult (retired tombstone, redirects here).
04 同仓库更多 Skills
clarify-requirements
Resolve genuine pre-execution ambiguity by scanning repo context, generating 1-6 grounded options, and asking 1-3 focused questions via AskUserQuestion. NOT for mid-task use (injection risk); NOT when the goal is already unambiguous; use to disambiguate vague prompts before any file edits.
openmm-md
Run FREE, open-source, Windows-native GPU molecular dynamics via the openmm-md MCP (OpenMM + CUDA on the local NVIDIA GPU). Use for protein / protein-ligand MD when you want a free stack or are on Windows without a Schrodinger license. Covers the full workflow (prepare, solvate, minimize, equilibrate, production, analyze, mmgbsa) AND an honest decision guide vs Schrodinger Desmond, including how to run Desmond legitimately on Linux or a cloud GPU with a real license. This is a deliberately LESSER free alternative and does NOT replace the schrodinger MCP tools. Do NOT use it to pick a free tool when a validated FEP+, WaterMap, IFD-MD, or Prime MM-GBSA result is actually required.
visual-review
Take screenshots of web pages and UI using Playwright for visual review and iteration
status
Diagnostic status report — shows metaswarm installation state, project setup, and potential issues
start
Use when starting work on any task, when the user mentions metaswarm, or when the user wants to begin tracked development work
setup
Interactive project setup — detects your project, configures metaswarm, writes project-local files