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/st0ff3l/fileterm/agents-mdgit clone --depth 1 https://github.com/St0ff3l/filetermWhat 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 | $0.03403 | $0.03403 |
| Opus 5 | $0.01702 | $0.01702 |
| Sonnet 5 | $0.00681 | $0.00681 |
| Haiku 4.5 | $0.00340 | $0.00340 |
Grade A, and why
fileterm 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 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 — 185 lines — stays where its author put it; the contents beside it link to each section on GitHub.
FileTerm Agent Guide
本文件是智能体进入 FileTerm 仓库时的入口地图，不是完整手册。详细事实以 docs/ 为准；当代码、设计或计划变化时，优先更新对应文档，不要把所有知识继续堆进这里。
1. 项目定位
FileTerm 是面向开发者与运维场景的 Rust + Tauri 桌面远程工作台，围绕 SSH / SFTP / FTP 构建可日常使用的多标签桌面客户端。apps/electron 仅保留为历史代码参考，不再参与构建、测试或发布。
当前阶段：Tauri 主链路稳定与发行收口。质量门禁覆盖共享包与 Rust/Tauri，Windows 使用签名的 Tauri 应用内更新；macOS 继续检查后跳转 GitHub Release 下载。
2. 先读哪里
- 架构地图：
docs/architecture.md - 设计规范：
docs/design.md - 路线图：
docs/roadmap.md - 进行中计划：
docs/plans/active/ - 已完成计划：
docs/plans/completed/ - 架构决策：
docs/decisions/ - 质量与回归：
docs/quality/ - 已隐藏功能：
docs/hidden-features.md - 功能草案：
.agents/extensions/ - 项目技能：
.agents/skills/
如果任务只改一个小点，先读本文件和相关源码即可；如果任务跨 Rust commands / bridge / renderer / packages 多层，必须先看 docs/architecture.md 和 docs/plans/active/。
3. 硬性边界
架构边界
packages/core是领域模型的 single source of truth。- Renderer 不直接访问 SSH / SFTP / FTP protocol clients。
- 所有系统能力必须走
Rust commands/events -> tauri-api.ts -> renderer。 - SSH/SFTP 与 FTP 在 controller/protocol 层保持分离，不做伪统一。
- Transfer 进度统一进入 Rust transfer service，不在组件里零散维护。
- 会话事件通过 Rust workspace runtime 的统一 event/channel 边界分发，不分散监听协议 worker。
- 新状态优先进入
packages/core定义类型，再下沉到服务层和 UI。 - 新窗口能力先定义 IPC 边界，再做 renderer 交互。
- 主题样式优先走
token -> theme vars -> component skins -> terminal colors。
平台兼容边界
- CWD 目录跟随：终端工作目录 (CWD) 变化通过底层会话流安全捕获，经 runtime 广播同步给文件管理器，严禁 UI 层轮询或直接探测平台路径。
- POSIX CWD 注入门控：
supportsPosixShellSetup()仅对linux/busybox返回 true。Windows / unknown 平台严禁注入 Linux shell CWD 脚本，采用 fail-closed 双重门控（detectPlatformAndSetupShell+injectShellSetup各一道）。 - CRLF 归一化：系统指标解析入口必须对远端输出做
replace(/\r\n?/g, '\n')归一化，避免'windows\r'等污染导致平台误判。 - Sudo 与 Root 状态同步：终端执行
sudo或切换用户态需被底层 runtime 解析，双向同步到文件管理器权限模型。
UI 与公用组件边界
- 语言名称使用本地自称：语言选择器中的语言名称必须保持该语言自己的写法（如
简体中文、English、한국어/조선어），不得根据当前界面语言翻译；新增语言时同样遵循此规则。 - 下拉框统一走 DropdownSelect：所有表单与设置项的下拉菜单必须统一使用公用组件
<DropdownSelect>，严禁直写原始 HTML<select>标签（确保 macOS 下包裹ft-select-shell外壳，Windows / Linux 下 100% 触发自绘 React Portal 弹出菜单）。 - 下拉箭头随控件缩放：
DropdownSelect的箭头必须由组件根据当前控件实际高度自适应（覆盖 macOS 原生外壳和 Windows/Linux 自绘触发器），业务组件不得写死一套箭头尺寸或覆盖共享计算；新增紧凑/表单尺寸时必须检查箭头与文字的垂直对齐。 - 图标矢量就地化：所有按钮与视觉图标优先使用预置的离线 SVG 图标组件
<AppIcon />，严禁新增<span className="material-symbols-outlined">依赖外部字体/WebFont 图标。 - 二次确认弹窗统一：所有破坏性/危险操作（如删除、清空等）必须调用项目通用的
<ConfirmActionDialog>确认弹窗组件，严禁在桌面 Webview 环境中使用原生window.confirm()。 - 按钮尺寸高度规范：同一操作组/表单行内的按钮必须具有严格统一的高度（如 32px 紧凑型 / 36px 表单型）、边框半径与内边距，禁止主次按钮尺寸参差不齐。
- 颜色语义边界：
--focus-outline只用于焦点/选中/拖拽目标的描边或光环；文件相关操作使用--folder-accent，实心主按钮使用--button-primary-*，不要用描边色填充按钮。 - 滚动条统一走公用组件：Renderer 中新增或改造的纵向滚动区域，默认必须复用
features/common/VerticalScrollbar.tsx，像终端区、文件区一样通过scrollRef绑定，并隐藏容器原生纵向滚动条；除非用户明确要求特殊行为，禁止在业务组件里单独绘制一套滚动条。横向滚动、第三方编辑器内部滚动和协议组件自带滚动可保留各自实现，但不得替代纵向公用滚动条。
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 · 185 lines · 3,403 tokens per session scan A 2677c366e71d
fileterm AGENTS.md is an instructions file published in the GitHub repository St0ff3l/fileterm (297 stars, last pushed 2d ago), licensed MIT. It adds 3,403 tokens to every session, about $0.0170 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
zenml CLAUDE.md
Claude Code instructions for zenml-io/zenml, covering zenml claude code guidelines, project structure, use zenml docs via mcp, code style & quality standards and commenting policy — explain why, not what.
zenml AGENTS.md
AGENTS.md instructions for zenml-io/zenml, covering zenml codex agent guidelines, project structure, always-loaded rules, common commands and branches, git, and prs.
livecodes AGENTS.md
Instructions for live-codes/livecodes, covering agents.md, project overview, documentation for ai agents, build / lint / test commands and project structure.
devopness AGENTS.md
Instructions for devopness/devopness, covering agents, core flow, shared constraints, monorepo structure and git workflow.
gpui-component CLAUDE.md
Claude Code instructions for longbridge/gpui-component, covering claude.md, canonical design and coding guides, project overview, common commands and development and testing.
aki-mcp-sv CLAUDE.md
Instructions for lacvietanh/aki-mcp-sv, covering aki-mcp-sv — project guidance, two client paths, one oauth server, os-agnostic by decision, not by accident, session lifecycle and process topology (stage 2 — single process).