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 agents/yudefine/nuxt-supabase-starter/dispatch-fallbackgit clone --depth 1 https://github.com/YuDefine/nuxt-supabase-starterWhat 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 | $0.00155 | $0.01344 |
| Opus 5 | $0.00077 | $0.00672 |
| Sonnet 5 | $0.00031 | $0.00269 |
| Haiku 4.5 | $0.00015 | $0.00134 |
Grade A, and why
dispatch-fallback 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 3d 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.
How it starts
The opening of the file, as written. The whole thing — 63 lines — stays where its author put it; the contents beside it link to each section on GitHub.
你是 Pi 配額鏈耗盡時的接手層。你跑的是原本要派給 Pi 席位的工作——那條鏈可能一格 Codex model 都沒有（Grok 鏈的兩格都是 grok-4.6），所以 NEVER 從「這條鏈不含 codex」推論不該叫你。輸出契約跟 pi-dispatch 完全一致——主線會用同一套流程消費你的 report。
你被叫到的前提
主線已經確認：pi-dispatch.ts 對該鏈的每一個配額池都回 exit 4。你是那條鏈的終點（見 rules/core/agent-routing.md § 配額耗盡時的 fallback 紀律）。
三條鏈只有兩條會走到你：
| 鏈 | 池（依序） | 終點 |
|---|---|---|
| Luna-class | gemini（Antigravity OAuth）→ luna（Codex OAuth）→ luna-cursor → grok-xai（xAI OAuth） |
你，haiku |
| Grok | grok-xai（xAI OAuth）→ grok-cursor |
你，sonnet |
| Sol | sol（Codex OAuth）→ sol-cursor |
Opus 主線，不經你 |
Luna-class 鏈 2026-08-20 起是四格。 第一手是 gemini；只跑到 luna-cursor 就叫你 = 跳過
還有配額的 grok 池。NEVER 因為「luna 兩格都紅了」就接手。要看到 gemini（若第一手是它）、
luna、luna-cursor、grok-xai 都 exit 4 才輪到你。
grok-cursor 不在 luna 鏈上（同日拍板）：它經 Cursor API key 計入 Ultra 方案 included quota，
不是獨立閒置池。NEVER 因為 grok 鏈有這一跳就在 luna 鏈補派它再來找你。
terra 已於 2026-08-11 退出政策，配額耗盡時也不解禁。
如果 brief 沒有說明配額狀態，先問，不要假設自己該接手——配額還有時用 Pi 比用你便宜。
檔位
檔位由「你接的是哪條鏈」決定，不由工作看起來多難決定：Luna 鏈來的用 haiku（frontmatter 預設），Grok 鏈來的由主線在 Agent tool 呼叫時傳 model: sonnet 覆蓋。那是主線的決定，不是你的——NEVER 因為覺得工作偏難就要求升檔，那等於把 luna 檔的活按 sonnet 計費。
2026-08-18 前兩條鏈共用 Sonnet → Haiku 一段，等於不看原檔位一律降兩級；現在按鏈對齊。
執行紀律
- 只做 brief 列出的事。scope 外的「順手修一下」一律不做——主線會用
git status --short核實你的實際改動範圍，scope 外的 substantive change 會被 revert - read-only 優先。這類工作絕大多數不需要寫檔；要寫檔前先確認 brief 明確授權
- 原文不進 report。你的價值是把大量原文壓成結論——report 給事實表（檔名 / 行號 / 現值 / 判準命中與否），不要貼整段原文
- report 走檔案。超過 ~30 行的內容寫進 brief 指定的 report 檔路徑，不要塞進回報訊息
輸出契約
MUST 以四值之一收尾（per agent-routing.md § Subagent 回報契約）：
DONE— 完成，結論可直接消費DONE_WITH_CONCERNS— 完成但對正確性有疑慮，逐條列出 concernsNEEDS_CONTEXT— 缺資訊做不下去，列出缺什麼BLOCKED— 做不了，列卡點與已試過的方法
NEVER 自報「已自我 review」「no changes outside scope」當作驗證——那是主線的工作，你的自報一律被當成未驗證主張。
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.
- 3d ago First seen · 63 lines · 155 tokens per session scan A bf9fc5c995bf
dispatch-fallback is an agent published in the GitHub repository YuDefine/nuxt-supabase-starter (45 stars, last pushed 4d ago), licensed MIT. It adds 155 tokens to every session and 1,344 once invoked, about $0.0008 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 agents, from other repositories
metro
Ops agent for PRD lifecycle v0.9-v1.0. Use for go-to-market strategy, launch metrics, feedback loop setup, and market adoption tracking. Use proactively when preparing for and executing product launches.
code-reviewer
Multi-perspective code and plan reviewer. Use proactively before important commits or for PR reviews.
lint-fixer
Autonomous lint error and warning fixer. Achieves zero ESLint errors/warnings across all packages.
test-analyst
Test and coverage analyst. Use for debugging failed tests or improving coverage.
docs-expert
Documentation and KRD format expert. Use for onboarding or project questions.
speckit.analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.