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 skills add AbyssCN/oh-my-dag --skill contractgit clone --depth 1 https://github.com/AbyssCN/oh-my-dagWrote 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/skills/abysscn/oh-my-dag/contract)<a href="https://agentmods.dev/skills/abysscn/oh-my-dag/contract"><img src="https://agentmods.dev/badge/skills/abysscn/oh-my-dag/contract/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/abysscn/oh-my-dag/contract"><img src="https://agentmods.dev/badge/skills/abysscn/oh-my-dag/contract.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00102 | $0.06603 |
| Opus 5.5 | $0.00041 | $0.02641 |
| Sonnet 5.5 | $0.00020 | $0.01321 |
| Haiku 4.5 | $0.00010 | $0.00660 |
Grade A, and why
contract 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.
How it starts
The opening of the file, as written. The whole thing — 289 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/contract — 审议结晶成执行契约
审议(/grill)收敛后,把结论写成结构化 SDD 落盘——它是执行方(Opus 子 agent 或 omd-lite run)的执行契约,写给没有对话上下文的执行器,不是给人读的散文。
承接 /grill 的决策记录表
| # | 决策 | 定论 | 落点(/note · SDD 章节) | 证据 |
|---|
落点标「SDD 章节」的行 → 进「决策」与「契约」段;标 /note 的留台账。「待 owner / 待实测」项原样进「未决」,不写成结论。
落盘
路径:docs/plan/YYYY-MM-DD-<slug>.md。每段可被无上下文执行器独立消费:
# <标题>
## 目标 (Destination) 一句话讲清做成什么样
## 决策 (Decisions) D-1..D-N:每条已定型裁决 + 一句为什么 + 证据
## 契约 (Contracts) 不变量 + GWT 验收点——执行方验收时逐条判 pass/fail 的依据
## 分解 (Breakdown) 依赖图 + 并行波形(写法见下)
## 非目标 (Non-goals) 明确不做什么
## 未决 (Open) 未裁的问题;待实测的标「待实测」
GWT 验收点越可证伪,验收越不含糊;一个模糊验收点 = 执行器和你各自解读的裂缝。
分解段 —— 默认并行,串行必须给理由
- 只写真实依赖边,每条带理由:「B 依赖 A 因 <B 消费 A 的产物 X>」——且 X 必须出现在 A 的写集清单里(机器可锚)。写不出被消费产物的边,删——那是顺序偏好,不是依赖。
- 接口依赖就地消解:B 只需要 A 的类型/签名/schema 而非实装 → 把该接口冻进本 SDD(契约段),边删掉,两片并行。接口从此是契约的一部分:A 的实装必须符合它,偏离 = 回流改契约,不是恢复边。(实例:内环 v2 切片 2 只消费切片 1 的
SddSlice结构——冻进 SDD 即可并行,当时没冻,白等一层。) - 定稿前过一遍并行性审问:串行率(关键路径长 ÷ 切片数,编译器
parallelismReadout同口径)= 1 的链,逐边重跑上面两问再定稿;审完仍是真串行就如实保留——读数只点名,不否决诚实的线性分解。 - 切片表:| 切片 | 写集 | 依赖(带理由) | verify |。写集两两不相交 = 可并行的机器判据。
- verify 列必须是可跑命令串(如
bun test src/x.test.ts),不是验收点引用——G 点引用的家在契约段。编译器(sdd-compile)按首词白名单拒不可跑的 verify;写引用 = 这份 SDD 吃不到直通 v2,回落 conductor 铺图税(owner 2026-08-11 裁:直通就是为了消铺图税与墙钟)。全量回归命令不进 verify 列(它只属终局 accept,D-4)。 - 收尾写并行波形一行,conductor 照铺:
并行波形:{1,3,5} → {2} → {4}。全串行波形须给整体理由,默认视为写错。 - 优先序(哪片价值高)写波形注释或 Open 段,不进依赖边。
⚠ 分解表的四条机器硬要求(写错 = 点火当场拒 / 静默回落,而它只在点火那一刻才响)
真源 src/harness/goal/sdd-direct.ts + sdd-compile.ts。2026-08-18 量过全量语料:
149 份 plan 文档、52 份有分解段,只有 14 份吃得下 sddPath 直通 —— 而直通是夜批的默认路径。
73% 的失败全部来自下面四条,一条都没写进过本模板:
- 切片列必须裸数字开头 ——
| 1 span 计算 …。写S1/**S1 …**/Phase一律拒 (sdd-direct.ts的/^(\d+)/;编号是波形/依赖引用它的唯一方式)。 - 波形行独占一行 ——
WAVE_LINE锚行首。写成「写集两两不相交 ✓。并行波形:…」会被 静默忽略(不报错,图退回按依赖边排;2026-08-18 实测)。 - 写集只收相对路径,且不许留空 —— 写「无」/「dispatch」这类词会被拒;空写集直接拒 (写集是并行安全的机器判据)。
- verify 列要过 #251 判据自证的三条 lint —— 真源
src/harness/goal/ignition-criteria-check.ts,点火期接线在goal.ts:1153。 这是六道点火闸里唯一没有force出口的一道:给了出口就等于放虚判据进图, 而虚判据让整跑判不出真假(实账 run85a18995零改动 4 分钟假 done;bca0a0c7的bun test多 filter 静默忽略缺失,二次假 done)。三条逐条写:
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 · 289 lines · 102 tokens per session scan A 1108fa38e6f7
contract is a skill published in the GitHub repository AbyssCN/oh-my-dag (43 stars, last pushed 5d ago), licensed MIT. It adds 102 tokens to every session and 6,603 once invoked, about $0.0004 per session on Opus 5.5. 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 2026-10-12.
Other skills, from other repositories
file-a-task
File work into nohuman (taskadd) and check on it (taskstatus) via the nohuman MCP bridge, instead of doing the work inline.
company-employee
The protocol every employee of a PenguinHarness organization follows — read the handbook first, act on [orgtrigger] work runs, schedule tickets from the desk session and do the work in ticket sessions, block instead of idling, ask the board in the all-hands channel before anything heavy, costly, irreversible or…
company-ceo
Run a PenguinHarness organization as its CEO — turn the mission into a ticket tree, hire HR and finance first, partition the shared workspace, schedule the calendar, open a channel per stream, review tickets and report to the board in the all-hands channel.
company-hr
Run HR for a PenguinHarness organization — guarantee every employee has an enabled calendar event, hire and offboard employees (channels included), evaluate and improve them (with the default-rsi-toolkit plugin's agent-optimization skill), and keep the handbook's role conventions current.
company-mirror
Run a PenguinHarness organization that mirrors a real company — one digital twin per real colleague, each twin's desk bound to that colleague's chat bot, twins answering what the handbook already knows and relaying the rest to one another instead of hiring, scheduling and filing tickets.
company-setup
Create a company-mode organization together with the user — collect its id, name, mission, shared workspace, model and CEO budget one question at a time, confirm the whole thing in one summary, then run penguin org create and hand the user over to company mode.