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 instructions/orziz/odai/agents-mdgit clone --depth 1 https://github.com/orziz/odaiWrote 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/instructions/orziz/odai/agents-md)<a href="https://agentmods.dev/instructions/orziz/odai/agents-md"><img src="https://agentmods.dev/badge/instructions/orziz/odai/agents-md.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 | $0.01852 | $0.01852 |
| Opus 5 | $0.00926 | $0.00926 |
| Sonnet 5 | $0.00370 | $0.00370 |
| Haiku 4.5 | $0.00185 | $0.00185 |
Grade A, and why
odai AGENTS.md 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 yesterday.
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 — 40 lines — stays where its author put it; the contents beside it link to each section on GitHub.
仓库维护约束
odai 必用
- 本仓库中的每次用户请求都必须使用 canonical
skills/odai/SKILL.md治理。若宿主明确声明当前odai-dsh-agent或odai-dsh-plugin已把完整 canonical governance 注入系统提示,并给出 skill version、runtime contract 与 bundle digest,则该注入本身视为已完整加载;声明有效且本会话未改动 canonical bundle 期间,不得再调用 skill 或read重复载入同一正文。仅当注入声明缺失或不可信、任务直接审查或修改 Odai、或 manifest 声明的 canonical 文件在注入后发生变化时,才须在行动前完整读取当前SKILL.md,并以新内容继续治理。其他宿主若当前任务尚未完整读取其当下版本,任何调查、判断、修改或测试前先完整读取。 - 按 odai 的真实缺口决定是否读取支撑资料;使用 odai 不等于自动增加计划、路由、角色、文件、测试或流程,简单任务仍直接完成。
- 对 odai 自身及其配套机制的修改同样受 odai 治理:先判断是否必要、是否比现状更成事、是否有可验证净增益,再决定保留、修改或退役,不因已有实现、历史方案、题本得分或局部可实现性继续堆叠。
- 以减少 token、延迟或成本为目的的优化,必须先用行为契约、反例与测试证明既有能力和安全边界不退化,再评估节省是否成立;不能验证保持项、只测预算下降、依赖模型自报或以删除能力换取指标时,不得合入或发布。
- 评测维护、题本设计、裁判与报告整理仍按上述要求治理;真正受测的 runner 则严格服从冻结的评测臂,不继承本节。
on只加载该臂声明的冻结 skill 与项目材料;off必须在干净隔离环境运行,不读取或注入 odai、.odai/local.md、odai 路由或 Hooks、要求使用 odai 的仓库指令、其他臂输出、既往 runner 转录或其派生状态。
官方 skills 单一事实源
skills/odai/与skills/ribao/是各自唯一可编辑的 canonical source;odai 是统一入口与最终交付 owner,ribao 是可独立加载的专业汇报能力。cli/skills/不在仓库中常驻;它只由 npmprepack临时生成,并在postpack清理。- 即使用户或 IDE 指向打包期间临时出现的
cli/skills/,也要把对应修改落到仓库根skills/<name>/。 - source 修改完成后,运行
node scripts/validate-odai-skill.mjs验证 canonical skills。 - canonical references 按单一所有权维护:
references/planning.md只负责正式计划、可执行合同与跨轮续作,references/craft.md只负责已决定结果的制作工艺,references/leverage.md只负责能力与责任选择、调度和交接;角色合同只引用对应 owner,不复制另一层的清单或宿主机制。 - odai-cli 未冻结时,发布相关修改还需运行
npm --prefix cli run pack:dry-run,确认产物与当前声明的打包范围一致,且命令结束后没有遗留cli/skills/;冻结期间遵循下节边界。
odai-cli 冻结边界
cli/、plans/odai-cli-plan.md与plans/odai-cli-runtime-canary.md当前冻结,不属于 active canonical / DSH 维护与验收范围。除非用户明确重新开放,不修改、测试、打包、升版或发布 odai-cli,也不为兼容它改变 canonical 或 DSH 路线。- 冻结期间,canonical / DSH 变更不运行 CLI tests 或
npm --prefix cli run pack:dry-run;只记录可能影响未来恢复的已知差异。cli/skills/仍不得作为可编辑 source 或遗留在仓库中。
版本发布约束
- 从本约束生效后的首个候选版本起,本仓库自有的任何新版本标识都不得包含数字字符
4;npm package version、skill version 与 runtime contract version 均受约束,版本序列遇到含4的候选值时必须直接跳过。已发布历史版本及其兼容记录必须按事实保留,不因本约束改写;上游依赖版本不属于本仓库自有版本,不受此规则限制。 - 禁止只靠人工记忆执行本约束。所有发布入口与版本事实源必须调用仓库的统一版本策略校验;新增版本载体时必须同时纳入该校验,命中禁用字符即 fail-closed。
- 同一 npm 候选从未发生过 registry publish,且未废弃或经用户明确拆分时,后续改动继续归入该候选;不得仅因新一轮任务、canonical / runtime contract 变化或新增 changelog 内容另起包版本,也不得把当前候选写成历史。版本一旦曾发布,即使后来 unpublish 或 deprecate,该标识也视为已消耗,不得复用。改版本前先核对 package metadata、当前唯一 Unreleased owner 与真实 registry 发布事实。
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.
- yesterday Changed · +6 lines · +286 tokens per session 394562efd580
- 5d ago First seen · 34 lines · 1,566 tokens per session scan A 53a3a9bf40a4
odai AGENTS.md is an instructions file published in the GitHub repository orziz/odai (107 stars, last pushed yesterday), licensed MIT. It adds 1,852 tokens to every session, about $0.0093 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 instructions, from other repositories
dsh-enhanced-plugins AGENTS.md
Instructions for sky-unicorn/dsh-enhanced-plugins, covering agents.md — deepseek harness 插件项目通用规则, 权威来源与开始工作前的检查, 插件架构与职责边界, cordis 插件与生命周期 and 配置与 settings.
dsh-virtual-product-team AGENTS.md
AGENTS.md instructions for songoao25/dsh-virtual-product-team, covering agents.md — 给 ai 编码助手/agent 的仓库说明, 这是什么, 仓库结构, 关键约定 and 常用命令.
deepseek-harness CLAUDE.md
Claude Code instructions for deepseek-ai/deepseek-harness, a project described as: DeepSeek Harness: Everything is a Plugin.
dsh-config-manager AGENTS.md
Instructions for xiajiajun516/dsh-config-manager, covering agents.md — dsh config manager 仓库协作指南, 🗣️ 语言与交流, 📦 项目概览, 🗂️ 仓库结构与职责 and ui 分层(改动前必读,避免放错层).
dsh-industry-research AGENTS.md
AGENTS.md instructions for PerryLink/dsh-industry-research, covering agents.md, layout, hard rules applied here, checks and release.
dsh-checkpoint-rewind AGENTS.md
AGENTS.md instructions for PerryLink/dsh-checkpoint-rewind, covering agents.md, 仓库布局:发布面 / 本地工程面, 命令, 提交纪律 and dsh 插件约束(红线).