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/wangbinquan/agent-workflow/claude-mdgit clone --depth 1 https://github.com/wangbinquan/agent-workflowWhat 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 | $0.10162 | $0.10162 |
| Opus 5 | $0.05081 | $0.05081 |
| Sonnet 5 | $0.02032 | $0.02032 |
| Haiku 4.5 | $0.01016 | $0.01016 |
Grade A, and why
agent-workflow CLAUDE.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 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.
How it starts
The opening of the file, as written. The whole thing — 272 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Repository status
v1 已发布：M0–M5 的 81 个 issue 全部完工（STATE.md §路线图全局视图：M0 5/5、M1 18/18、M2 16/16、M3 14/14、M4 11/11、M5 12/12），发布产物由 v* tag 触发的 workflow 产出。此后所有产品 / 技术工作一律以 RFC 形式落地（design/RFC-NNN-{slug}/），见 §RFC workflow。
Read in this order at session start:
STATE.md— session-to-session execution log. Always read first; tells you what's done, what's next, current caveats.design/plan.md— RFC 索引（编号 / 标题 / 状态；新 RFC 在此登记）+ 已完工的 M0–M5 路线图存档。这里已无待认领 issue，新工作从 RFC 开始。design/RFC-294-backend-layered-target-architecture/— 后台全局目标架构总纲；新 RFC 的设计必须朝它演进（见 §RFC workflow 第 8 条）。design/proposal.md— product spec (authoritative).design/design.md— technical design (authoritative).proposal/init.md— original Chinese proposal, preserved for history. When it disagrees withdesign/*.md,design/*.mdwins.docs/dev-gotchas.md— 跨 RFC 沉淀的通用踩坑（提交纪律 / 迁移 / CI / opencode / impl-gate 经验规律 / 前端 / dev-env）。动手前扫一遍，避免重复踩坑；踩到新的通用坑也补进去（RFC-专属细节仍进各design/RFC-XXX/）。
When a batch of work (RFC tasks, fixes) completes, commit + push and update STATE.md so the next session can pick up seamlessly.
bun install 装依赖。质量 / 测试门禁以 GitHub Actions 为准，本地不再要求跑（2026-08-24
用户明令，详见 §Test-with-every-change 的「运行门槛」）：bun run gate:local（完整本地门禁）与
bun run test（backend + shared + frontend）保留为可选的诊断入口——用于复现 CI 报出的红，
不是提交的前置条件。
RFC workflow（新增 / 修改前的强制流程）
任何超出 design/plan.md 已列 issue 范围的新功能、非平凡重构、产品行为变更，必须先走 RFC，再写代码：
- 落档：在
design/RFC-NNN-{slug}/子目录下创建三件套proposal.md—— 产品视角：背景、目标 / 非目标、用户故事、验收标准design.md—— 技术设计：接口契约、数据流、与现有模块的耦合点、失败模式、测试策略plan.md—— 任务分解：编号子任务（RFC-NNN-T1...）、依赖、PR 拆分建议、验收清单
- 编号：递增分配，从
RFC-001起；在design/plan.md的 "RFC 索引" 表里登记新条目（标题 + 状态：Draft / In Progress / Done / Superseded）。 - 用户确认：RFC 写完后必须用
ExitPlanMode或显式询问得到用户批准，才能进入实现阶段。不要边写 RFC 边改代码。 - STATE.md 同步：RFC 落档同时在
STATE.md顶部追加一行"进行中 RFC"指向新目录。RFC 完工后把状态改为 Done 并在STATE.md已完成 issue 表里加一行（与 P-X-XX 同等级）。 - PR：单个 RFC 默认对应单个 PR，commit message 前缀写明
feat(scope): RFC-NNN 标题；如确实需要拆分，在plan.md里说明并分别立 PR。 - 不走 RFC 的例外：拼写 / 单行 bug 修复、纯重命名、依赖升级、文档增删、测试补充、CI 微调。这些可以直接改 + 提交。
- 能力收缩型 RFC 的附加门槛（RFC-224 事故沉淀）：凡以安全 / 隔离 / 密封为由关闭或收缩既有能力（含「新路径不再继承旧路径的能力」）的 RFC：
proposal.md必须含**「能力影响清单」**章节：逐项列出被关闭的既有能力与受影响的部署形态，作为 breaking change 呈用户逐项确认——不得以「安全默认」名义静默移除（RFC-224 曾静默切断自定义 provider 网关部署，生产无预警全挂，事后只能以 RFC-251 / RFC-255 逐个受控恢复）；- 每条禁用 / 拒绝分支必须有测试覆盖（禁用分支与正向功能同等对待，见
docs/dev-gotchas.md对应教训）； - 关闭判据必须是可复跑的外部源码引用（
file:line），接手复核规则同docs/dev-gotchas.md§「RFC / design 里对 opencode 行为的既有断言」。
- 目标架构对齐（RFC-294 总纲，强制）：
design/RFC-294-backend-layered-target-architecture/定义了后台的全局目标架构——feature-first bounded context + 模块内domain / application / engine / ports / infrastructure分层，执行链固定为 TaskEngine → WrapperRuntime → NodeExecutor → ExecutionKernel，跨模块只依赖 exactpublic/{commands,queries,participants,events,types}合同，bootstrap 唯一装配。此后每个新 RFC 都必须考虑向该架构做出架构演进：- 写
design.md前先读 RFC-294 的proposal.md §1 摘要裁决 / §3 目标与design.md，在设计里写明本次改动落在哪个 bounded context、哪一层，新增代码按目标架构落位； - 不要再往
routes//services/横向平铺层加新的跨域耦合、facade 或 cross-context 内部 import；顺手能把触及的存量结构朝目标架构挪一步就挪，并在design.md里写清「本 RFC 承担哪一步演进、留下哪些债」； - 确有偏离（必须绕过 kernel、必须新增临时 facade 等）时在
design.md里逐条列出偏离项与理由并呈用户确认，不得默默沿用旧形状； - RFC-294 本身是总纲、零生产改动：各演进波次仍各自立 RFC 单独获批，新 RFC 不因「对齐 294」就自动取得实现许可。
- 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.
- 2d ago First seen · 272 lines · 10,162 tokens per session scan A c1e250ad2aec
agent-workflow CLAUDE.md is an instructions file published in the GitHub repository wangbinquan/agent-workflow (23 stars, last pushed 2d ago), licensed Apache-2.0. It adds 10,162 tokens to every session, about $0.0508 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
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).
next.js AGENTS.md
Instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.