- name
- voice-reply-mode
- description
- Add same-modality conversation behavior to an agent: voice in → voice out, text in → text out. Use when enabling Telegram/Feishu voice-note workflows with Edge TTS, documenting agent voice preferences, and providing gateway config snippets for inbound-only TTS auto replies.
Voice Reply Mode
Use this skill when a user wants an agent to behave like this:
- 用户发语音 → agent 回语音
- 用户发文字 → agent 回文字
What this skill includes
- Workspace-level behavior rules for
SOUL.md/IDENTITY.md/TOOLS.md - A reusable Edge TTS helper script
- Gateway config snippets for
messages.tts - Channel notes for Telegram / Feishu style deployments
Important boundary
This skill can package rules, scripts, and config snippets.
It does not automatically change a user's global gateway config unless the user explicitly asks and authorizes it.
Recommended workflow
- Read
references/workspace-snippets.md - Read
references/gateway-config.md - Copy the relevant snippets into the target agent workspace
- If the user explicitly asks, patch gateway config with the snippet from
references/gateway-config.md - Validate with one text message and one voice message
Minimal success criteria
- Text message receives text reply
- Voice message receives voice reply
- Agent workspace documents the preferred voice
Files in this skill
references/workspace-snippets.md— snippets forIDENTITY.md,SOUL.md,TOOLS.mdreferences/gateway-config.md—messages.ttsexamples and caveatsreferences/channel-notes.md— Telegram / Feishu notesscripts/edge-tts.sh— helper script for local TTS generation
Notes
- In many deployments, the decisive switch is gateway-level
messages.tts.auto = "inbound". - Workspace files define behavior expectations, but gateway config determines whether automatic voice replies actually happen.
- If schema rejects
identity.voice, keep voice preference in workspace docs instead.