midasheng-audio-tagging

Audio tagging service for environmental sound recognition. Use when user needs to identify environmental sounds in audio files (water sounds, snoring, etc.)...

Security Scan

The skill's code and instructions match its stated purpose (audio tagging) and there are no unexplained credentials or installs, but it will upload user audio to a remote third‑party endpoint so review privacy before use.

Purpose & Capability

Name, description, SKILL.md, and the included script all describe and implement an audio-tagging client that uploads audio to a remote tagging API and checks queue metrics. The required capabilities align with the stated purpose.

Instruction Scope

The instructions and script are narrowly scoped to two actions: POST a multipart file to the tagging endpoint and POST to the metrics endpoint. However, both actions transmit user audio (or contact the service) over the network to https://llmplus.ai.xiaomi.com; the SKILL.md and code do not warn about privacy/PII implications. No other local files, credentials, or unrelated system state are read.

Install Mechanism

This is an instruction-only skill with a single Python script and no install spec. Nothing is downloaded or installed by the skill itself. The only runtime dependency is the 'requests' Python library (not declared), which is a minor packaging mismatch but not a security problem.

Credentials

The skill requests no environment variables or credentials (proportional). The main risk is data exposure: it sends user audio files to an external service (Xiaomi domain). Users should evaluate whether they are comfortable uploading potentially sensitive audio to that endpoint and confirm the service's privacy/legal policies.

Persistence & Privilege

The skill does not request persistent or privileged presence (always=false). It does not modify other skills or system configuration and does not require special privileges.

Assessment

This skill appears to be what it says: a client that uploads audio files to a remote tagging API. Before installing or using it, consider the following: - Privacy: Using the skill will upload audio files to https://llmplus.ai.xiaomi.com. Do not upload audio containing sensitive personal data, private conversations, or recordings you do not own or have permission to share. - Trust & policy: Verify who runs the endpoint and review their privacy policy and retention practices if you plan to send real data. - Local testing: Test with non-sensitive sample audio first to confirm behavior and outputs. - Dependency: The script uses the Python 'requests' library; ensure your runtime environment has it installed. - Alternate endpoints: The script allows overriding the API URL; if you have an on-prem or trusted tagging service, pass that URL instead to avoid sending data to the default remote service. If you need the skill to run without sending data externally, consider using or requesting a version that performs tagging locally or points to a self-hosted inference endpoint.

dasheng-audio-tag

Audio tagging service that identifies environmental sounds in audio files.

Quick Start

Tag an audio file:

python3 scripts/audiotag.py <audio_file>
# 服务地址: https://llmplus.ai.xiaomi.com/dasheng/audio/tag

Check queue status:

python3 scripts/audiotag.py --queue

API Details

Tagging Endpoint: POST https://llmplus.ai.xiaomi.com/dasheng/audio/tag

• Method: POST, multipart form-data with file upload

Queue Metrics Endpoint: POST https://llmplus.ai.xiaomi.com/metrics?path=/dasheng/audio/tag

• Returns: active（当前活跃请求数）、avg_latency_ms（平均耗时）
• 预估等待时长 = active × avg_latency_ms

Usage Patterns

Basic tagging

python3 scripts/audiotag.py audio.mp3

Check queue / waiting status

python3 scripts/audiotag.py --queue

Programmatic usage

from scripts.audiotag import tag_audio, check_queue, format_queue_status

# Tag audio
result = tag_audio("sample-0.mp3")

# Check queue
queue_info = check_queue()
print(format_queue_status(queue_info))

Queue Status（排队情况）

何时调用

1. IM 即将超时但 tag 服务还未返回结果时：调用 --queue 查排队情况，告知用户当前排队状态并请用户稍后来问任务是否完成。
2. 用户稍后询问任务进度但 tag 服务仍未返回时：调用 --queue 返回最新排队情况给用户。

返回字段说明

字段	说明
active	当前活跃请求数（排队中）
avg_latency_ms	平均处理耗时（毫秒）
estimated_wait_sec	预估等待时长（秒）= active × avg_latency_ms
total_processed	累计已处理请求数

状态分级

• 🟢 active=0 或预估等待 <5s → 服务空闲/很快完成
• 🟡 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

Supported Audio Formats

Common audio formats: mp3, wav, m4a, ogg, flac

Troubleshooting

• File not found: Check the audio file path
• API request failed: Verify network connectivity and API endpoint availability
• Unsupported format: Try converting to mp3 or wav format
• Long wait: Use --queue to check current queue status