midasheng-audio-denoise

Voice enhancement and noise reduction service. Accepts a noisy audio file and returns a clean, denoised version. Use when user needs to remove background noi...

Security Scan

The skill's code and instructions align with its stated purpose (upload audio to a denoising API and return cleaned audio), but it sends user audio to an external Xiaomi-hosted endpoint and has minor inconsistencies (no dependency declaration for Python requests, SKILL.md lists curl). Review privacy and trust before use.

Purpose & Capability

Name, description, SKILL.md, and the included script consistently implement an audio denoise client that uploads a file to https://llmplus.ai.xiaomi.com/dasheng/audio/denoise and returns a WAV. The requested resources (no env vars, no unusual binaries) are proportionate to the described task.

Instruction Scope

Instructions and the script only perform network calls to the Xiaomi endpoint and a metrics endpoint, and read the provided input audio file and write an output WAV. They do not access other system files or credentials. However, the skill requires uploading user audio to a third-party service (privacy risk) — this is expected for a cloud denoise service but should be considered by the user.

Install Mechanism

No install spec; this is instruction-only with a bundled Python script. Nothing is downloaded or executed automatically beyond what the user runs. Low installation risk.

Credentials

The skill declares no environment variables or credentials, which is proportionate. Minor inconsistency: SKILL.md top lists 'requirements: curl' and examples use curl, but the provided script uses Python's requests library — requests is not declared as a dependency and may not be present in runtime.

Persistence & Privilege

Skill does not request always:true or modify agent/system configs. It is user-invocable and uses normal agent invocation permissions.

Assessment

This skill uploads the user's audio to https://llmplus.ai.xiaomi.com (a Xiaomi-hosted API) for processing. If your audio contains sensitive or private information, do not use the skill unless you trust that endpoint and its privacy policy. Also note the script uses Python's 'requests' but the SKILL.md mentions curl — ensure your environment has Python and requests installed before running, or use the curl example. If you need local-only processing for privacy, prefer a local denoising tool or verify the service's data retention and handling policies. Finally, because the skill's source/homepage is absent, consider testing with non-sensitive samples first and verifying the endpoint ownership if you require higher assurance.

Voice enhancement and noise reduction service powered by advanced backend algorithms.

1. Trigger

Use this skill when the user wants to:

• Remove background noise from an audio file
• Clean up a noisy recording
• Enhance voice clarity
• Preprocess audio for speech recognition

2. API Details

Endpoint: POST https://llmplus.ai.xiaomi.com/dasheng/audio/denoise (multipart/form-data)

Parameters:

• file: The audio file to denoise

Response: Binary audio stream (WAV format)

3. Usage

Basic denoising

curl -X POST "https://llmplus.ai.xiaomi.com/dasheng/audio/denoise" \
  -F "file=@noisy_recording.mp3" \
  -o clean_recording.wav

Script usage

python3 scripts/denoise.py noisy_audio.mp3 -o clean_audio.wav
python3 scripts/denoise.py --queue   # Check queue status

4. Queue Status（排队情况）

查询命令

python3 scripts/denoise.py --queue
# 或直接调 API：
curl -X POST "https://llmplus.ai.xiaomi.com/metrics?path=/dasheng/audio/denoise"

返回字段

• active: 当前活跃请求数
• avg_latency_ms: 平均处理耗时（毫秒）
• 预估等待时长 = active × avg_latency_ms

何时调用

1. IM 即将超时但 denoise 服务还未返回结果时：查排队情况告知用户，请用户稍后来问。
2. 用户稍后询问任务进度但服务仍未返回时：查最新排队情况返回给用户。

状态分级

• 🟢 active=0 或预估等待 <5s → 服务空闲
• 🟡 预估等待 5-30s → 轻微排队
• 🔴 预估等待 >30s → 排队较长，建议稍后重试

5. Supported Audio Formats

Input: mp3, wav, flac, ogg, m4a. Output: WAV.

6. Troubleshooting

• Empty output: Input file may be too short
• API request failed: Verify network connectivity
• Poor results: Works best on speech with background noise