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 skills add bleakbelladonnals/asr-transcription-skills --skill course-transcribegit clone --depth 1 https://github.com/bleakbelladonnals/asr-transcription-skillsWrote 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/skills/bleakbelladonnals/asr-transcription-skills/course-transcribe)<a href="https://agentmods.dev/skills/bleakbelladonnals/asr-transcription-skills/course-transcribe"><img src="https://agentmods.dev/badge/skills/bleakbelladonnals/asr-transcription-skills/course-transcribe/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/bleakbelladonnals/asr-transcription-skills/course-transcribe"><img src="https://agentmods.dev/badge/skills/bleakbelladonnals/asr-transcription-skills/course-transcribe.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00121 | $0.04833 |
| Opus 5 | $0.00060 | $0.02416 |
| Sonnet 5 | $0.00024 | $0.00967 |
| Haiku 4.5 | $0.00012 | $0.00483 |
Grade A, and why
course-transcribe scanned grade A with 1 finding 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 12d 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
- **排查方法（curl A/B 对照）**：同 key 换模型发请求——换模型后正常 = 模型级限流；全 429 = key 级问题 How it starts
The opening of the file, as written. The whole thing — 166 lines — stays where its author put it; the contents beside it link to each section on GitHub.
课堂录音转文字（course-transcribe）
一条命令完成「课程录音 → 高质量中文转写文本」，术语准确优先，成本趋近 0（免费额度）。
触发条件
- 用户提到：课程录音转写 / 录音转文字 / 把录音转成文本 / 转写后做复习蓝本
- 输入一般是录音豆导出的 OGG/MP3（存于
~/Documents/luyinwenjian/<日期>/）
技术栈（2026-08 实测定案，勿改）
| 环节 | 工具 | 角色 |
|---|---|---|
| 采集 | 飞书安可录音豆（硬件） | 录音（0 断录已验证；弃用钉钉A1录音卡 102 次断录） |
| 主转写 | Gemini gemini-3.6-flash |
免费层主力，15 分钟/段，约 40-80 分钟转完 5 小时课 |
| 备选/校对 | 百炼 qwen-audio-3.0-asr-flash |
5 分钟/段（服务端上限 300s），5 小时课约 20-30 分钟 |
| 热词 | 百炼即时热词 / Gemini prompt 术语表 | 免费，生僻术语纠偏 |
⚠️ 模型选型实测结论（2026-08）：百炼不要用旧模型 qwen3-asr-flash——它无热词（预编译被服务端拒绝
preprocess service not found、即时热词参数被静默忽略）、术语常错（Claude Code → Cloud Code）。qwen-audio-3.0-asr-flash全面更优：模型更强 + 热词生效 + 响应自带句子级毫秒时间戳。gemini-2.5-flash 对 2026 年新用户已下线（404），勿用教程旧模型名。
快速开始
# 1. 前置：key 文件（一次性）
# scripts/ 目录下放 gemini_key.txt（Google AI Studio 获取，需代理）和 bailian_key.txt（sk- 开头）
# 或用 --key-file / --gemini-key-file / --bailian-key-file 指定已有 key
# 2. 主转写（默认 Gemini，5 小时课约 1 小时）
python3 transcribe.py "~/Documents/转写评测/8.7/录音.ogg"
# 3. 备选/交叉校对
python3 transcribe.py 录音.ogg --engine bailian # 百炼单独转
python3 transcribe.py 录音.ogg --engine compare # 双引擎 + 锚点术语对比报告
完整工作流
第 1 步：录音采集（硬件，课前）
- 录音豆放讲台/离声源 1-2 米（独立设备，不依赖手机位置）
- 课后导出音频（OGG/MP3 均可，脚本自动处理）
- 脚本内置断录检测（分帧 RMS：0.5s 帧，相邻帧骤降 >20dB 计数）——检测到疑似断录会警告，漏录内容转写救不回，宁可重录
第 2 步：音频预处理（脚本自动）
- ffmpeg 统一转 16kHz 单声道 mp3 64kbps（ASR 标准输入）
- 分段：Gemini 15 分钟/段（规避免费层 TPM/请求体限制）；百炼 5 分钟/段（qwen-audio-3.0-asr-flash 服务端硬上限 300s，超了报
AUDIO_DURATION_TOO_LONG） - 分段串行调用（免费层并发会互相限流）
第 1.5 步：噪音段裁剪（可选，转写前先问用户录音是否有大段空场）
- 工具：
~/Documents/工具脚本/noise_handler.py（零依赖：标准库+ffmpeg） - 四种动作：
analyze音量分析给阈值建议 /detect只列低音量段时间戳 /cut -o 输出裁剪拼接 /silence -o 输出静音保时长 - 绝对静音法：
-t -35 -d 2（默认），适合有课间休息/无人声空场的录音 - 相对峰值法：
-r 25（低于峰值 25dB 算低音量段），适合无绝对静音但主讲人（近场）与讨论声（远场）音量差大的录音——先analyze看峰值，再定相对值 - ⚠️ 2026-08 实测：8.13 上午/下午 ogg 全程 mean -27~-34dB 无静音段，-r 25 检出 177 段全是 <25s 句间停顿（主讲人几乎全程在说话），无大段空场 → 不需要裁剪。先跑 detect 确认有没有可裁剪段再动手，别盲裁
- ⚠️ 若噪音是嘈杂人声（下课喧哗/讨论，音量不低），能量法失效 → 用 trim_smart.py（Gemini 预听定位，见下）
- ⚠️ brew ffmpeg 8.1 无 libvorbis、原生 vorbis/opus 编码器标记 experimental 报错——ogg 输出统一用
-c:a libopus
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 12d ago First seen · 166 lines · 121 tokens per session scan A f9b3083fb525
course-transcribe is a skill published in the GitHub repository bleakbelladonnals/asr-transcription-skills (9 stars, last pushed 26d ago), licensed MIT. It adds 121 tokens to every session and 4,833 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
stage-dsl
The map for reading and editing an OpenMAIC stage document with readstage, patchstage, and grepstage. Load it before patching a structure you have not patched before, when patchstage rejects an operation, or whenever the path from a stage, outline, scene, content object, or action to the field you need is uncertain.…
pptx-import
A method for adding an uploaded PowerPoint presentation to an existing classroom as extra pages while keeping the slides' original layout. PowerPoint is Microsoft's presentation file format.
dev-camp-deck
Creates a professional PowerPoint presentation about a Copilot Dev Camp lab or topic. Researches Microsoft Learn documentation, organizes content into a slide structure, and generates a polished deck with speaker notes. Use when user asks to "create a presentation on", "make a deck about", "build a slideshow for"…
read-book
When you want to read and extract structured notes from a book — PDF, EPUB, MOBI, markdown, .txt, pasted text, or URL to a public-domain work. Reads in chunks (by chapter when a TOC exists, by 50-page blocks otherwise), extracts per-chapter TL;DR + key concepts + quotes + action items + frameworks, and offers to…
exam-study-guide
A study-guide builder for a completed exam chapter that has not yet passed its required checks. It creates a structured teaching checklist and, in full mode, a self-contained HTML and printable PDF guide with readable formulas, visible images, explanations, examples, and answers.
training-report
Produce a professional training/workshop report as a .docx file. Use this skill whenever the user mentions "training report", "workshop report", "compte rendu", "compte rendu de formation", "formation report", "debriefing a workshop", "write up a training session", "résumé de formation", or any request to document a…