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/zkwi/voxtype/agents-mdgit clone --depth 1 https://github.com/zkwi/VoxTypeWrote 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/zkwi/voxtype/agents-md)<a href="https://agentmods.dev/instructions/zkwi/voxtype/agents-md"><img src="https://agentmods.dev/badge/instructions/zkwi/voxtype/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 | $0.02502 | $0.02502 |
| Opus 5 | $0.01251 | $0.01251 |
| Sonnet 5 | $0.00500 | $0.00500 |
| Haiku 4.5 | $0.00250 | $0.00250 |
Grade A, and why
VoxType 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 4d 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 — 218 lines — stays where its author put it; the contents beside it link to each section on GitHub.
VoxType AI 维护指南
本文件是 VoxType 的 AI 编码助手工作规则。 所有 AI 修改代码前,必须先阅读本文件。
VoxType 是一个 Windows 桌面语音输入工具。核心链路是:
触发录音 → 麦克风采集 → 豆包流式 ASR → 可选 LLM 润色 → 写入剪贴板 → 自动粘贴 → 恢复剪贴板 → 统计与日志
项目目标:实用、轻量、易维护。避免过度抽象、过度重构和不必要依赖。
1. 不可破坏的主链路规则
以下规则不得回退:
- 空识别必须进入失败态,不得显示“已粘贴”,不得触发润色、粘贴或成功统计。
- 只有 LLM 已启用、文本长度达到
min_chars,且 Base URL、API Key、模型名完整时,才显示“正在润色文本”。 - 悬浮字幕只显示实时字幕、耗时状态和错误提示;不得显示“正在粘贴”“已粘贴”等瞬时状态。
- 写入剪贴板失败时,必须给用户兜底提示;如果文本已复制,应提示用户可手动
Ctrl + V。 - 统计不得记录识别正文,只能记录时长、字数、速度等非正文数据。
- 日志和诊断报告默认不得包含真实密钥、识别正文、热词、prompt、最近上下文正文或 Windows 用户名路径。
- 最终输出必须等待豆包最终包;实时字幕和中间包只作反馈,不得进入润色、粘贴、最近上下文、自动热词历史或成功统计。
definite=true二遍分句用于稳定最终文本,但不能压掉更完整且高度重合的最终包整段文本;最终包补齐开头或尾字时应优先保留最终包。- 识别速度优化优先通过首字加速、OCR 超时折中、响应轮询和字幕节流实现,不得为了体感速度把默认 ASR 音频分片改离豆包建议的
100-200ms范围,默认200ms不应轻易缩短。 - 尾字真实被切断时,优先调整停录尾音窗口和 flush/负包链路,不得用中间 ASR 文本兜底成功。
ASR 速度、识别效果、OCR 等待、尾音和最终包选择相关改动,还必须参考 docs/asr-quality-latency-guardrails.md。
2. 不可回退的默认策略
以下默认值不得改回激进状态:
- 最近上下文默认关闭。
- 右 Alt 触发默认关闭。
- 鼠标中键触发默认关闭。
- 录音期间静音系统声音默认关闭。
- 最终识别正文默认不打印到控制台。
- 自动粘贴后默认尝试恢复原剪贴板。
3. 代码修改原则
- 一个任务只解决一个问题。
- 不要顺手重构无关文件。
- 不要在同一次修改中同时改 ASR、剪贴板、热键、UI 大结构和配置结构。
- 不要新增没有必要的依赖。
- 不要把用户可见文案写死在业务逻辑中,应优先放到 i18n。
- 不要在首页展示内部路径、协议细节、debug 状态或内部错误堆栈。
- 修改配置项时必须同步 Rust 默认值、配置模板、前端设置项、三语言文案和 README。
- 修改用户可见行为时必须同步 README 或相关 docs。
- 修改用户配置、功能说明、排障流程或发布流程时,应同步检查 README、README.en.md、Wiki 和
docs/wiki/草稿。 - 修改主链路时必须说明影响范围和手工验证步骤。
- 若不确定,应先输出方案,不要直接改代码。
4. 目录职责
前端
src/routes/+page.svelte:当前主页面入口。可以小范围修改,但不要继续塞大型新功能。src/lib/components/:新增 Svelte 组件应放这里。src/lib/i18n/:三语言文案和错误码文案。src/lib/:可放前端纯函数、格式化函数、状态映射函数。
Rust / Tauri
src-tauri/src/session.rs:录音会话状态机。src-tauri/src/audio.rs:麦克风采集。src-tauri/src/asr.rs:ASR 请求组装、上下文和结果解析。src-tauri/src/asr_ws/:豆包 WebSocket ASR 会话、音频发送、最终文本和错误映射。src-tauri/src/llm_post_edit.rs:LLM 润色。src-tauri/src/text_output.rs:剪贴板写入和模拟粘贴。src-tauri/src/hotkey.rs:全局热键、右 Alt、鼠标中键输入钩子。src-tauri/src/config.rs:配置模型、加载、保存、校验。src-tauri/src/app_log.rs:日志与脱敏。src-tauri/src/stats.rs:非正文统计。src-tauri/src/overlay.rs:悬浮字幕。src-tauri/src/tray.rs:系统托盘。
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.
- 4d ago First seen · 218 lines · 2,502 tokens per session scan A fec9f67f4424
VoxType AGENTS.md is an instructions file published in the GitHub repository zkwi/VoxType (88 stars, last pushed 10d ago), licensed MIT. It adds 2,502 tokens to every session, about $0.0125 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
kesha-voice-kit AGENTS.md
Instructions for drakulavich/kesha-voice-kit: Project rules, build commands, release process, and lessons learned live in CLAUDE.md — single source of truth for every coding agent (Claude Code, Cursor, Aider, Codex, etc.). When CLAUDE.md and any other doc disagree, CLAUDE.md wins.
blurt AGENTS.md
Instructions for satyaborg/blurt, covering agents.md, what is blurt, commands, dev install and run.
vzt-flow CLAUDE.md
Instructions for vonzelle-vzt/vzt-flow, covering claude.md — vzt flow, workspace layout, build / test / run, critical gotchas and verification norms.
kesha-voice-kit CLAUDE.md
Instructions for drakulavich/kesha-voice-kit, covering claude.md, project overview, critical development rules, default tts voices must be male and never auto-download the engine or models.
blurt CLAUDE.md
Instructions for satyaborg/blurt, a project described as: Talk to your coding agents.
voicemode CLAUDE.md
Claude Code instructions for mbailey/voicemode, covering claude.md, voice interaction, project overview, key commands and development & testing.