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/starcat-app/starcat/agents-mdgit clone --depth 1 https://github.com/starcat-app/StarcatWrote 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/starcat-app/starcat/agents-md)<a href="https://agentmods.dev/instructions/starcat-app/starcat/agents-md"><img src="https://agentmods.dev/badge/instructions/starcat-app/starcat/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 | $0.07439 | $0.07439 |
| Opus 5 | $0.03719 | $0.03719 |
| Sonnet 5 | $0.01488 | $0.01488 |
| Haiku 4.5 | $0.00744 | $0.00744 |
Grade A, and why
Starcat 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 today.
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 — 449 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
本文档为跨 Agent 协作提供指导，确保多个 Agent 在本代码库工作时保持一致。
🚨 硬性铁律（每次写代码前必读，违反即返工）
⚠️ 【铁律 #1】本项目已发布正式版，有线上用户与本地数据。 ⚠️ 已随正式版发出的数据库 schema：变更必须追加
registerVN迁移，禁止直接改已落地的v1-initial，禁止要求用户删库重建。 ⚠️ 未发布功能开发期例外：可直接改该功能建表草稿 SQL +ensurePrelaunch*补齐本机库；收口进正式版时压成单次registerVN并收口。依赖未建表的中间 migration 必须对「表不存在」no-op。 ⚠️ RAG 已于 2026-07-14 收口为v7-knowledge-rag：禁止再回写 v1 草稿或启动期旁路补 RAG。 ⚠️ 应用层仍禁止堆叠无意义的「兼容旧 API / 双写旧字段 / 永久双轨」——废弃路径用迁移删掉或一次性替换，不要长期保留两套接口。 ⚠️ 看到代码层旧路径不再需要 → 直接删；已发布库的字段废弃 → 追加迁移处理，不要只改建表语句假装老用户会跟着变。
⚠️ 【铁律 #2】方案讨论 ≠ 动手许可。 ⚠️ dong4j 在反馈方案 / 修正理解 / 给出补充信息时，默认仍在讨论阶段。 ⚠️ 必须等 dong4j 明确说「开干 / 改吧 / GO / 动手 / 实施」等字眼，才能开始改代码；只要还在交换意见就只读不写。
⚠️ 【铁律 #3】禁止擅自执行打包 / 发布 / 上传脚本。 ⚠️ 除非 dong4j 在当前消息里明确要求执行，否则 Agent 只能修改脚本、写文档、给命令，不能执行
scripts/package-*、scripts/release-*、deploy.sh、notary 上传、App Store 上传、服务器上传等会生成或发布分发产物的命令。 ⚠️ 允许执行只读检查命令，例如bash -n、git diff --check、codesign -d、PlistBuddy -c Print、dwarfdump --uuid。
⚠️ 【铁律 #4】禁止擅自改写
docs/功能实现总览.md。 ⚠️ dong4j 未口头 / 书面明确确认前，禁止对该文件做任何写入：含勾选[x]、追加 / 改写> 实现：、变更日志、进度仪表盘数字、新增- [ ]/ 技术债条目、改状态说明、删改正文、任何措辞润色。 ⚠️ 改完代码或其它文档 ≠ 可以同步改总览。 中间迭代、未验收、未确认的改动，一律不准往变更日志里塞条目；禁止连刷返工垃圾日志。 ⚠️ 允许且鼓励只读打开该文件做开工检查。 需要登记时：在回复里起草拟写入内容，等 dong4j 明确说「可以写总览 / 同步总览 / 记到总览 / 勾上」等之后再改文件。 ⚠️ dong4j 确认后写入时仍须遵守下文勾选与> 实现：格式（仅打勾不够）。
🌿 Git 分支与 Worktree（强制）
所有分支与 worktree 的创建、切换、同步、合并和删除，必须遵循 docs/5-规范/Git-分支与Worktree规范.md。
- 涉及分支/worktree 操作前，先读取根目录
BRANCH.md，再用 Git 命令核对真实 refs、提交关系和 worktree 归属。 - 创建、切换或恢复长期分支后，必须向 dong4j 说明当前分支、用途、基线、worktree、状态和后续归宿。
- 创建、合并、停放、废弃或删除长期分支/worktree 时，必须在同一任务中同步更新
BRANCH.md。 - 删除前必须检查独有提交、Git 合并关系、语义覆盖和 worktree 脏状态；未审查分支禁止直接删除。
- 未经 dong4j 明确授权，不得 push、删除远端分支、强制删除 worktree 或改写共享历史。
BRANCH.md 记录分支意图；Git refs 与 git worktree list 是实际状态的事实来源。
🧭 主进度索引（每次开工前必读）
docs/功能实现总览.md 是本项目的【活文档主索引】，所有 P0/P1/P2 功能与重构债务都在那里以 checkbox 形式记录。
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.
- today Changed · +20 lines · +362 tokens per session 167294861d28
- 4d ago First seen · 429 lines · 7,077 tokens per session scan A 5ce57a3833c7
Starcat AGENTS.md is an instructions file published in the GitHub repository starcat-app/Starcat (132 stars, last pushed yesterday), licensed MIT. It adds 7,439 tokens to every session, about $0.0372 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
local-rag-core CLAUDE.md
Claude Code instructions for linkwut-create/local-rag-core, covering claude.md — local-rag-core, mcp integration, required pre-modification workflow, architecture boundaries and key commands.
siyuan AGENTS.md
AGENTS.md instructions for siyuan-note/siyuan, covering agents.md, 1. non-negotiable constraints, do not hand-edit, verification and prohibited operations and 2. project-specific rules.
personal-model AGENTS.md
Instructions for Intuition-Lab/personal-model, covering agents.md, runtime boundary, commands, pipeline and documentation map.
humla CLAUDE.md
Claude Code instructions for michaelwilhelmsen/humla, covering humla — project notes, what this app is, core capabilities, architecture overview and data flow during a recording.
personal-model CLAUDE.md
Instructions for Intuition-Lab/personal-model: Read and follow AGENTS.md. It is the single agent orientation for this repository: Runtime boundary, commands, architecture, invariants, current documentation, privacy rules, and DCO workflow.
openmessage CLAUDE.md
Instructions for MaxGhenis/openmessage, covering openmessage, architecture, local cli (read-only, no transports), multi-platform import and mcp serving modes.