1. 权限边界
AI agent 将继承你的账户权限。
Agent 可以创建转写任务、消耗积分，并读取你的 Audio2SRT.app 账户或 API Key 允许访问的结果。
只对你信任的 Agent 使用。原始 API Key 和 bearer token 应保存在密钥存储中，不要放在聊天里。
公开 Skill 文件只保存工作流规则；你的凭据保留在自己的运行时。
Skill 链接
https://audio2srt.app/skills/audio2srt-agent/SKILL.mdAI agent 通道
让 Agent 使用你的 Audio2SRT.app 账户权限。
2. Agent 交接
复制给你的 Agent
Load and follow this Audio2SRT.app skill before transcribing media:
https://audio2srt.app/skills/audio2srt-agent/SKILL.md
Permission boundary:
- You are acting as my authorized Audio2SRT.app agent.
- You inherit my Audio2SRT.app account permissions for this task.
- You may create transcription jobs, spend my Audio2SRT.app credits, and read task results allowed by my account or API key.
- Do not print, log, or expose raw API keys, bearer tokens, uploaded media URLs, or private transcript content unless I explicitly ask.
Credential handling:
- Store AUDIO2SRT_API_KEY in your secret manager.
- Never paste the raw key into chat or generated files.
- Call Audio2SRT.app from server-side or trusted agent runtime code only.
- Official site: https://audio2srt.app
Runtime configuration:
AUDIO2SRT_API_BASE=https://asr.xiaojins.com
AUDIO2SRT_API_KEY=<store in your secret manager>
Workflow:
1. Submit audio/video media with POST /v1/transcribe/upload, or public video links with POST /v1/transcribe/link.
2. Save the returned task_id.
3. Poll GET /v1/tasks/{task_id} until the status is completed or failed.
4. Return available transcript layers: punctuated text, SRT, word timestamps, and speaker segments.
5. If the task fails, report the failure reason and do not retry indefinitely.