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/1139030773-cmd/agent-workflow-system/claude-mdgit clone --depth 1 https://github.com/1139030773-cmd/agent-workflow-systemWrote 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/1139030773-cmd/agent-workflow-system/claude-md)<a href="https://agentmods.dev/instructions/1139030773-cmd/agent-workflow-system/claude-md"><img src="https://agentmods.dev/badge/instructions/1139030773-cmd/agent-workflow-system/claude-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 5.1 | $0.02576 | $0.02576 |
| Opus 5 | $0.01288 | $0.01288 |
| Sonnet 5 | $0.00515 | $0.00515 |
| Haiku 4.5 | $0.00258 | $0.00258 |
Grade A, and why
agent-workflow-system 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 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 — 187 lines — stays where its author put it; the contents beside it link to each section on GitHub.
⚡ 输出前自检（每次回复前强制执行，不可跳过）
| 1 | 话题层面 | 这轮在聊什么范畴？ | | 2 | 话题变了吗 | 变了 → 必须先话题标定再回复 | | 3 | 要写文件吗 | 写了什么文件？层面匹配吗？（工作流文件 ≠ 开发内容） | | 4 | 重复建设？ | completed 里有没有已经做过类似的事？ | | 5 | 有依据吗？ | 关于系统的设计判断 → 必须有源文件引用。纯推理不允许，先去读再回来说 |
启动恢复检查
只有以下情况才扫描 .resume/ 走恢复流程：
| 用户第一句话 | 行为 |
|---|---|
| "继续" / "接着" / "上次" / "恢复" | 扫描 .resume/ 目录 → 按活跃 session 数量处理（见下方） |
| 沉默 / 明确新任务 | 跳过，不打扰 |
多 session 选择逻辑
扫描 .resume/ 下所有 session-*.md 文件，过滤 status=active
0 个活跃 session → fallback 读根目录 RESUME.md（向后兼容）
1 个活跃 session → 直接按 5 字段格式展示，问「接下来想怎么做？」
≥2 个活跃 session → 列出编号让用户选：
| # | 任务 | 阶段 | 最后活跃 |
|---|------|------|----------|
| ① | {task_name} | {phase} | {last_updated} |
| ② | {task_name} | {phase} | {last_updated} |
回复 1/2 选择，或者说"新任务"跳过
用户选定后再按 5 字段格式展示，问「接下来想怎么做？」
Session 文件格式
文件路径：.resume/session-YYYYMMDD-HHmmss.md（时间戳 = 收尾时刻）
每个文件结构与根目录 RESUME.md 一致，额外包含 session_id 字段：
- session_id: session-YYYYMMDD-HHmmss
- status: active | paused | completed
- task_name: ...
- phase: ...
- last_updated: ...
- completed: ...
- next_step: ...
- blocked_by: ...
- task_stack: ...
- context_snapshot: ...
- last_session_end: ...
收尾写文件规则
收尾时写两个位置：
.resume/{session-id}.md— 该窗口专属 checkpoint（主）- 根目录
RESUME.md— 向后兼容 fallback（副，只保留最新一个任务的信息）
同一任务多次收尾用不同 session ID → 自然产生多个文件 → "继续"时按 task_name 去重，只展示每条任务的最新文件。 |
版本更新检查
SessionStart hook 已自动检查远程版本号并写入 .claude/.version-check。第一轮回复时读这个文件即可。
文件格式：newer|远程版本|本地版本（只有远程 > 本地时才存在）
如果存在 → 提醒用户。不存在 → 版本一致，沉默。 无需手动 WebFetch。
系统健康度检查
每次会话启动时，静默运行 scripts/health-check.ps1 check（或 .sh）。
全部健康（≥90%）→ 沉默，不打扰用户。 有问题（<90%）→ 在第一轮回复末尾附加报告。
检测范围：RESUME 完整性、session 文件字段、TASK_QUEUE 损坏/孤儿引用、task_stack 孤儿任务、DECISIONS 完整性、Skill 组件、膨胀检测。
同脚本 clean 模式：自动归档旧 session 文件、旧 completed 条目、标记旧 decisions。只归档不删除。
系统改动铁律
只要改了系统文件，以下 8 步必须一口气做完，不等用户催：
| # | 步骤 | 说明 |
|---|---|---|
| 1 | 四目录同步 | ① 根目录 c:\Users\11390\Documents\New project\（CHANGELOG/README.md/README_CN.md/CLAUDE.md）② plugins/agent-workflow-system/（插件 CLAUDE.md 规则）③ plugins/plugins/agent-workflow-system/（plugin.json/skills）④ agent-workflow-system-local/（GitHub 展示用 README/CHANGELOG/CLAUDE.md — 用户看到的页面来自这个仓库） |
| 2 | CHANGELOG | 在对应版本条目下记录改动（Added/Changed/Fixed），版本号必须与步骤 5 的 plugin.json 一致 |
| 3 | README | 如果改动影响功能描述，同步更新 README.md + README_CN.md 两份文件 |
| 4 | RESUME.md + .resume/ | 如果改动涉及任务状态变更，同步更新 .resume/{session-id}.md + 根目录 RESUME.md（last_updated / completed / next_step / phase） |
| 5 | 版本号 | 如果是新功能/breaking change，同步更新 plugin.json + marketplace.json 版本，版本号必须与步骤 2 的 CHANGELOG 一致 |
| 6 | 对抗验证 | spawn 独立 agent，读取本次改动的所有文件，专门找"应该改但没改"的遗漏（四目录是否都覆盖了？CHANGELOG 写了但 README 漏了？版本号改了但 marketplace.json 没改？）。发现遗漏 → 回去补，不准跳过。无遗漏 → 继续 |
| 7 | commit + push + release | 在 plugins/ 和 agent-workflow-system 仓库提交推送；打版本 tag；gh release create 创建 GitHub Release（--latest 标记最新版本） |
| 8 | 版本一致性验证 | 运行 scripts/validate-version.ps1（或 .sh），通过（显示 ✅）才算完成。不通过 → 回去补步骤 2 或 5，不准跳过 |
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 · 187 lines · 2,576 tokens per session scan A 4add41010021
agent-workflow-system CLAUDE.md is an instructions file published in the GitHub repository 1139030773-cmd/agent-workflow-system (5 stars, last pushed 2mo ago), licensed MIT. It adds 2,576 tokens to every session, about $0.0129 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 instructions, from other repositories
scream-code AGENTS.md
AGENTS.md instructions for LIUTod/scream-code, covering scream-code development guide, workspace overview, packages, terminology and cross-package import rules.
spare-me CLAUDE.md
Instructions for RazorEdge-wyh/spare-me, covering claude.md, what this project is, common commands, directory layout and engineering conventions.
MindOS CLAUDE.md
Claude Code instructions for GeminiLight/MindOS, a project described as: MindOS is a Human-AI Collaborative Mind System, where human thinks and agents act. Globally sync your mind for all agents: transparent, controllable, and evolving symbiotically.
ReMe AGENTS.md
AGENTS.md instructions for agentscope-ai/ReMe, covering agents.md, project principles, sources of truth, repository map and development setup.
open-swe CLAUDE.md
Claude Code instructions for langchain-ai/open-swe, a project described as: An Open-Source Asynchronous Coding Agent.
cyrus copilot-instructions.md
Copilot instructions for cyrusagents/cyrus: Note, there is a need to maintain the use of '--print' when running the claude exec commands because that is what makes it non-interactive.