Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/lee-w/maigo/model-dispatchnpx skills add Lee-W/maigo --skill model-dispatchgit clone --depth 1 https://github.com/Lee-W/maigoWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00112 | $0.01047 |
| Opus 5 | $0.00056 | $0.00524 |
| Sonnet 5 | $0.00022 | $0.00209 |
| Haiku 4.5 | $0.00011 | $0.00105 |
Grade A, and why
model-dispatch scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
Model Dispatch
Owner: orchestrator Consumers: maigo orchestrator，在 Claude Code harness 下透過 Agent tool spawn subagent 時
只在 Claude Code harness 下適用——Agent tool 的 model 參數能讓 orchestrator 對每次
spawn 個別 override 檔位；沒有這個能力的 harness（例如 Codex 的 command-router 環境）
不適用本 skill，交辦一律走該 harness 原生的模型設定。
檔位政策
| 檔位 | 用途 |
|---|---|
haiku |
機械批次：格式轉換、read-back、逐檔套用已定型的修法 |
sonnet |
預設工作馬：搜尋、實作、重構、審查、研究——多數交辦都用這檔 |
opus |
升級檔位：卡關救援、高風險判斷的第二意見 |
Orchestrator 本身執行所在的模型若高於以上三檔，只用在規劃、制度設計、品味判斷這類需要 「取捨」的工作——不實作、不掃 repo、不批次改檔。凡是可交辦的機械或搜尋工作，一律往 下派給對應檔位的 subagent，不因為自己在高檔位上就順手代勞。
升降級
- haiku 錯 1 次 → 同任務升
sonnet重派。 - sonnet 同一子任務連錯 2 次 → 升
opus，且必須帶完整失敗軌跡（原始交辦 prompt、 前兩次的輸出與錯誤原文）——不是重講一次題目給 opus 猜。 - opus 解出模式後 → 降回
sonnet/haiku批次套用到其餘案例。「解出模式」的判準： 解法能寫成 ≤5 步、不需再做判斷、其他檔位可照抄執行的固定步驟；寫不成就還沒解出， 不要降級。
重試預算
同一件事最多重試兩輪，計數明確定義：初次交辦不算重試；重試第 1 輪＝原檔位修一次；
重試第 2 輪＝升級後（opus）修一次。opus 輪在兩輪預算之內；opus 輪仍失敗即停手——
判斷是方向錯（換路）還是需要使用者輸入（去問），不再無聲重試第三輪。
SendMessage 續用與換檔位
SendMessage 續用既有 agent 時不能換 model；要換檔位就開新 agent、附上前情
（原交辦內容＋已知結果），不能讓新 agent 從零猜任務背景。
不升檔的情況
Reviewer 已指定具體修法、且修法已機械化（可照抄執行、不需再判斷）時，續用原檔位即可—— 不因為「上一輪出過錯」就自動升檔；升檔只在「同一子任務連錯」時觸發，不是每次失敗都升。
與既有 skill 的分工
「要不要派 subagent」與「派了之後怎麼交辦、怎麼防止 orchestrator 失焦」屬於
skills/harness-discipline
的範圍；本 skill 只管「派給哪個檔位」。兩者搭配使用：先用 harness-discipline 的門檻判斷
要不要派，再用本 skill 判斷派給誰。
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 60 lines · 0 tokens per session scan A 8dc258475e69
model-dispatch is a skill published in the GitHub repository Lee-W/maigo (14 stars, last pushed 9d ago), licensed MIT. It adds 112 tokens to every session and 1,047 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
agent-messaging
Send and receive cryptographically signed messages between AI agents using the Agent Messaging Protocol (AMP). Use when the user asks to "send a message to an agent", "check agent inbox", "message another agent", "reply to a message", "notify an agent", or any inter-agent communication task.
kagent-dev
Development guide for kagent's v1alpha3 Harness and AgentTemplate CRDs, AgentInstance gRPC control plane, upstream A2A integration, Substrate runtime provisioning, tests, generation, and PR workflow. Use for any implementation, debugging, review, or CI task in the kagent repository.
workflows
Execute a workflow script that orchestrates multiple subagents deterministically. Use when the user explicitly requests a workflow, multi-agent orchestration, parallel or sequential agent pipelines, structured agent outputs, or a durable background workflow run.
agent-tuning
Use when changing agent model or effort configuration, adapter mappings, or eval candidate profiles.
ax-python-agent
Use when writing Python code with axllm for agents, child delegation, tools, MCP, citations, persistent playbook learning, stage instructions, runtime state, final typed responses, and direct-respond executor skipping.
📝 任务完成后归档
重要提醒: 每次完成复杂调试或开发任务后，主动执行此流程！ 将学到的经验归档为 skill，供以后参考。不要等用户提醒。.