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 asr-transcription-evaluationgit 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/asr-transcription-evaluation)<a href="https://agentmods.dev/skills/bleakbelladonnals/asr-transcription-skills/asr-transcription-evaluation"><img src="https://agentmods.dev/badge/skills/bleakbelladonnals/asr-transcription-skills/asr-transcription-evaluation/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/asr-transcription-evaluation"><img src="https://agentmods.dev/badge/skills/bleakbelladonnals/asr-transcription-skills/asr-transcription-evaluation.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 | $0.00106 | $0.01543 |
| Opus 5 | $0.00053 | $0.00772 |
| Sonnet 5 | $0.00021 | $0.00309 |
| Haiku 4.5 | $0.00011 | $0.00154 |
Grade A, and why
asr-transcription-evaluation 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 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.
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 — 49 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ASR 转写与评测（Speech-to-Text: Transcription & Evaluation）
何时使用
- 把录音（课程/会议/访谈）转成文本或 SRT
- 对比多家转写引擎或录音硬件，输出评测结论/文章
- 搭建可复用的多平台转写管线（API Key 文件 + 脚本）
- 判断某段录音/某台设备是否有漏录断录
核心原则（血泪教训）
- 控制变量：CER 等量化对比必须用同一份音频做金标准。两台设备同场录制 → 时间轴会漂移（前段错位 4-5 分钟、后段对齐），对比要用内容锚点对齐，不能按时间戳对齐。
- 客观指标必须结合人耳听感：动态范围大/有静音段 ≠ "信号干净"，可能是断录/漏录。本次实测：钉钉录音卡 102 次电平骤降 + 最长 11 秒漏录，被客观指标误读为"动态自然"，用户一句听感反馈推翻结论。动态大 ≠ 好，连续才是硬道理。
- 统一输入格式：全部 ffmpeg 转 16kHz 单声道 mp3（64kbps）分段上传——一步规避 OGG 兼容性（讯飞不认 OGG）和各平台大小限制。
- 模型有生命周期：文档示例模型会过期（gemini-2.5-flash 对 2026 年新用户 404 "no longer available"）。先用用户的 key 查 models 列表，再选模型；控制台 > 文档 > 记忆。
- 免费层串行 > 并行：两任务并行抢额度互相限流，重试堆积（曾 16 分钟跑不完 12 段）。
平台速查（详细坑位见 references/provider-pitfalls.md）
| 平台 | 调用方式 | 关键限制 |
|---|---|---|
| Gemini | POST /v1beta/models/{model}:generateContent + inline_data base64 |
模型名会过期；先查 /v1beta/models |
| 阿里百炼 | 必须用 dashscope SDK（OpenAI 兼容模式只收公网 URL，6.29MB 请求体上限） | SDK 同步接口单段 ≤240s（用 3 分钟段）；SDK 返回 dict 不是对象 |
| 火山方舟 | /api/v3/chat/completions + input_audio base64 |
必须先在控制台"开通管理"激活模型（否则 ModelNotOpen 404） |
| OpenAI | /v1/audio/transcriptions |
25MB/文件，长音频切段 |
评测维度（5 层 20 项，详见 references/benchmark-methodology.md）
- 第 0 层 采集可靠性（硬件必测，20%）：断录事件数/最长低电平段/低电平帧占比/主观听感
- 第 1 层 识别准确性（30%）：CER/术语锚点/数字/信息完整率
- 第 2 层 鲁棒性（20%）：噪声/口音/中英混杂/说话人分离
- 第 3 层 语义可用性（20%）：语义保真/标点/幻觉率/格式/口语还原
- 第 4 层 工程商业（10%）：成本/时延/门槛
脚本（scripts/）
transcribe_multi.py— Gemini/火山 通用转写（16kHz 切段 → 分段调用 → 拼接）bailian_sdk_transcribe.py— 百炼专用（dashscope SDK，3 分钟段）benchmark_eval.py— 金标准对比评分：语义保真/CER(参考)/幻觉/完整率/锚点 → 控制台表 + HTML 报告；自动适配整理式听写金标准（语义级权重 40/25/20/15）audio_dropout_check.py— 断录检测：0.5s 帧 RMS 能量分析（骤降计数/最长低电平段/低电平占比）
陷阱速查
- 金标准类型决定评测方法：人工听译多交付"整理式听写稿"（理顺口语/改词），字级 CER 会全引擎虚高（实测 30-67%）→ 改语义级评分（句匹配相似度），且注意它偏向措辞简化型引擎（通义语义分虚高但锚点垫底）。给转录者的指令必须写"逐字、保留嗯啊重复"；报告注明金标准类型（详见 references/benchmark-methodology.md）
- 新版 dashscope SDK 返回 dict：
resp.output["choices"][0]["message"]["content"][0]["text"]——用.text属性访问会 AttributeError，每段失败重试空转 ~2.5 分钟（本次最大时间坑） - 幻觉检测：hyp 中与 gold 无匹配的连续 insert 块（>15 字）累计字数占比
- 通义听悟免费版：幻觉（凭空英文/脏话误听）+ 信息丢失（蜜雪冰城/GitHub 缺失）+ 导出非标准 SRT（无时间戳）——不可用于复习蓝本；结论不适用于付费版/百炼 API
- 百炼 enable_itn=True 输出阿拉伯数字（35%），False 输出中文（百分之三十五）；火山 doubao 固定输出中文数字——跨引擎对比注意格式差异
- 锚点类对比要区分"引擎问题"和"音频问题"：同一引擎换音频后锚点表现骤变 → 音频质量差异（Gemini 的 Claude Code 识别 0 错→8 错）
What ships with it
6 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 · 49 lines · 106 tokens per session scan A 2c9af59b5ae0
asr-transcription-evaluation is a skill published in the GitHub repository bleakbelladonnals/asr-transcription-skills (9 stars, last pushed 26d ago), licensed MIT. It adds 106 tokens to every session and 1,543 once invoked, about $0.0005 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 skills, from other repositories
arboreto
Infer gene regulatory networks (GRNs) from gene expression data using scalable algorithms (GRNBoost2, GENIE3). Use when analyzing transcriptomics data (bulk RNA-seq, single-cell RNA-seq) to identify transcription factor-target gene relationships and regulatory interactions. Supports distributed computation for…
pyhealth
Build clinical/healthcare deep-learning pipelines with PyHealth — loading EHR/signal/imaging datasets (MIMIC-III/IV, eICU, OMOP, SleepEDF, ChestXray14, EHRShot), defining tasks (mortality, readmission, length-of-stay, drug recommendation, sleep staging, ICD coding, EEG events), instantiating models (Transformer…
torchdrug
Build and troubleshoot TorchDrug 0.2.1 workflows for molecular graphs, property prediction, self-supervised pretraining, molecule generation, retrosynthesis, protein representation learning, and knowledge graph reasoning. Use when code imports torchdrug or needs its datasets, models, tasks, or Engine.
deepspot-m
Generate transcriptome-wide virtual spatial transcriptomics from H&E histology with DeepSpot-M. Use when you need spatial gene expression in log1p-CPM for 224x224 tiles at about 20x, want to query protein-coding genes by symbol instead of a fixed panel, or want to run prediction across a whole slide after tiling with…
nemo-mbridge-perf-expert-parallel-overlap
Validate and use MoE expert-parallel communication overlap in Megatron-Bridge, including overlapmoeexpertparallelcomm, delaywgradcompute, and flex dispatcher backends such as DeepEP and HybridEP.
pick-a-pii-model
Select an on-device OpenMed PII model from the committed registry by language, runtime format, and size budget, then require recall validation before deployment. Use when an agent must choose a local PII detector for CPU, Apple Silicon, or a mobile export without relying on live model discovery.