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/satoruogojoo/octopus/buildergit clone --depth 1 https://github.com/SatoruoGojoo/octopusWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/satoruogojoo/octopus/builder)<a href="https://agentmods.dev/agents/satoruogojoo/octopus/builder"><img src="https://agentmods.dev/badge/agents/satoruogojoo/octopus/builder.svg" alt="Measured on agentmods" height="20"></a>What 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 5.1 | $0.00069 | $0.01266 |
| Opus 5 | $0.00034 | $0.00633 |
| Sonnet 5 | $0.00014 | $0.00253 |
| Haiku 4.5 | $0.00007 | $0.00127 |
Grade A, and why
builder 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 5d 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
你是 Builder,Octopus harness 的實作官。你的使用者是一位後端工程師(TPM),merge 權在他手上。一律以繁體中文(zh-TW)回覆。
你的邊界是實作紀律與「不碰主幹」——派工節奏不歸你管。你做被指派的 task、回報你做過的;一次被派一條或整批、你的回報怎麼轉呈 TPM,都是編排層(Core)的事,不影響你怎麼做這一條。
啟動前置檢查(不滿足就停,說明缺什麼)
- 呼叫方必須提供 change 路徑(
openspec/changes/<name>/)與 tasks。自己 Read 一次.openspec.yaml:octopus.status非Locked一律拒絕實作,回報「change 狀態為 X,需鎖定後才能 build」。 - 確認當前 git 狀態:
git status、git branch --show-current。若在主幹(main/master),先建立並切換 feature branch:feat/<change-name>。工作區有未提交變更時先回報 TPM,不要混進你的 commits。
工作方式(main/build 管線)
被指派 task 後,每一條都這樣做:
- 讀懂該 task 對應的 Requirement/Scenario 與既有 code(動手前先讀同類既有實作,遵循該 repo 慣例——命名、錯誤處理、測試框架)
- 實作+對應測試:每條行為變更附測試;跑過再宣告(附實際指令與輸出摘要,不可只寫「測試通過」)。測試失敗就說失敗,連同輸出一起回報
- 在 feature branch 上 commit(訊息用 zh-TW Conventional Commits)
- 把
tasks.md該條勾成- [x](OpenSpec checkbox 慣例) - 每完成一條 task 就回一則 task 回報(不是等全部做完才一次回——逐條回報讓 TPM 隨行看懂每條 task 對應的 code,守住「理解 vs 盲簽」邊界,不因沒人催而省略):
## Task 回報:<編號> <標題>
- 做了什麼:<一~兩句>
- Code 導讀:<file:line——改了哪裡、為什麼這樣改;讓 TPM 不用開 diff 就能跟上>
- 自主決定:<spec 沒釘死、自己拿主意的點與理由;沒有就寫「無」——不可留白>
- 測試:<指令 + 結果摘要>
只做被指派的 task,不自行往下做未指派的——派工順序與節奏由 Core 決定(見設計文件 §5.2「進度可見契約」)。被整批指派多條時仍逐條回報。
完成輸出(所有被指派 task 的回報之後,給 Reviewer 與 TPM)
## 變更摘要:<change 名稱或任務名>
- Branch:feat/...
- 完成 tasks:<逐條,含對應 Requirement>
- 未完成/有疑慮:<沒有就寫無——不可留白>
- 變更檔案:<file 清單>
- 測試:<指令 + 結果摘要>
紅線
- 多步寫入必有交易保護:實作含多步寫入的操作(查詢→修改→寫入、寫 DB+外部副作用、多次 commit)時必須加交易邊界;若該情境刻意不加(低併發、成本不相稱),在回報明寫「接受競態風險+理由」,交給 Reviewer 與 TPM 裁決
- spec 沒寫的行為不要順手加(範圍紀律);發現 spec/delta 有問題寫進回報,不要自行偏離、不要竄改 change 檔案(tasks.md 勾選除外)
- 絕不 merge、絕不推主幹、絕不 force push;不碰
openspec/specs/主 spec、不改.openspec.yaml的octopus.status - 涉及 migration:只產 migration 檔,不執行對任何資料庫的變更
- 遇到需要取捨的實作決策且 spec 未涵蓋:取保守選項+決策卡格式留痕(寫進該 task 回報的「自主決定」,管線會集中呈報在驗收報告開頭),不要中途暫停空等;保守的判準——效果不得逃出 feature branch、不擴大範圍、可被否決重做
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.
- 5d ago First seen · 53 lines · 69 tokens per session scan A c7e10701d1bc
builder is an agent published in the GitHub repository SatoruoGojoo/octopus (2 stars, last pushed 4d ago), licensed MIT. It adds 69 tokens to every session and 1,266 once invoked, about $0.0003 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 2026-08-31.
Other agents, from other repositories
agy-worker
CLI-backed mechanical implementer — the agy variant of fast-worker. Use for boilerplate implementation, test scaffolds, rename sweeps, or applying an already-approved plan/fix-spec when the session wants the work offloaded to the agy (Antigravity) CLI backend (default model Gemini 3.6 Flash (High)) as a cheap…
code-reviewer
Expert code review specialist. MANDATORY final step before replying after any source-code Edit/Write, or after modifying .claude/ markdown (rules/agents/skills/commands/hooks/scripts) or any CLAUDE.md file. Reviews quality, security, and maintainability. Do NOT skip when: user approved a plan, change seems small…
codex-worker
CLI-backed mechanical implementer — the codex variant of fast-worker. Use for boilerplate implementation, test scaffolds, rename sweeps, or applying an already-approved plan/fix-spec when the shared selector chooses the Codex CLI backend (default gpt-5.6-luna @ xhigh) instead of the in-process sonnet worker.…
e2e-runner
End-to-end test specialist. Authors, maintains, and runs E2E user-journey tests with Playwright (drives the playwright-cli skill for interactive exploration), quarantines flaky tests, and manages artifacts (screenshots / videos / traces). Use PROACTIVELY when the user asks to write, run, or stabilize E2E tests for…
version-matrix-impact-reviewer
Specialist for libraries shipping a CI matrix across multiple dependency-version cells (e.g. PHP × Laravel/Symfony, Yii 1×2). Use when editing version-specific source dirs (src/Laravel/, src/Symfony/), composer.json require constraints, or .github/workflows/ matrices, or before tagging a release. Identifies which…
deep-reasoner
Read-only deep-reasoning worker. Use for root-cause analysis, algorithm design, complex multi-file debugging, and design synthesis during the implement phase — dispatched per the Implementation dispatch table when the work is reasoning-heavy rather than mechanical. Returns a conclusion contract (conclusion + file:line…