360 Secure SkillHub
360 安全技能中心

SoundClaw

该技能面向 OpenClaw,用于首次接入并操作 SoundClaw 音频后端。它检查 soundclawctl 是否可用,并以 runtime status 判断就绪状态;随后通过运行时 CLI 解析资产与输出、控制播放与停止、调整音量、执行场景或 Layer,并进行健康与配置检查。它只中继运行时结果,不负责后端安装、修复或回滚,也不解析 Web UI 或主机文件。

catholicbeer
catholicbeer
热度 13.51
最近更新 2026-07-20 23:53:09
分类:自动化依赖:代码执行 / 读写文件应用场景:自动化脚本 / 工作流 / DevOps

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add soundclaw
Metadata
name
soundclaw
description
Onboard and operate SoundClaw through one thin OpenClaw-facing product and the promoted runtime CLI.
version
1.0.0
homepage
https://github.com/catholicbeer/soundclaw-skills/tree/main/skills/soundclaw
metadata
{"openclaw":{"os":["linux"]}}

SoundClaw

Use this skill for every SoundClaw operator request. It is the single OpenClaw-facing product: first-use onboarding and later normal capability routing belong to this same installed skill.

First-Run And Readiness Behavior

  • When backend readiness has not been established in the current conversation, check for soundclawctl on PATH; if it is absent, check /opt/soundclaw/runtime/current/bin/soundclawctl.
  • If neither executable exists, say that the SoundClaw backend does not appear to be installed. Point only to the public GitHub Releases page at https://github.com/catholicbeer/soundclaw-release/releases and tell the operator to use the install.sh wrapper inside one verified public soundclaw-pi-release-<release-id>.tar.gz bundle. Stop there. Do not fetch, install, repair, copy, or sync backend content.
  • If an executable exists, run <soundclawctl> runtime status --json. Treat that runtime-owned result as authoritative. Report ready, degraded, or unavailable without attempting host repair.
  • Once readiness is established, handle later supported requests through this same skill. Do not require or route to separately installed SoundClaw leaf skills.

Capability Routing

  • Resolve assets with soundclawctl library list [--json] and soundclawctl library show --asset <id> [--json].
  • Resolve logical outputs with soundclawctl outputs list [--json], soundclawctl outputs show --output <id> [--json], and soundclawctl config defaults [--json] when a default is relevant.
  • Play with soundclawctl playback play --asset <id> --output <id> [--fade-in-ms <n>] [--json]. Stop with soundclawctl playback stop --output <id> [--fade-out-ms <n>] [--json].
  • Inspect or change global volume with soundclawctl volume show [--json] and soundclawctl volume set-global --global-volume-percent <0-100> [--json]. Explain any runtime-reported reload requirement; never hide a reload.
  • Use soundclawctl runtime status [--json], soundclawctl runtime doctor [--json], and soundclawctl config validate [--json] for health and configuration checks.
  • Report backend deployment identity with soundclawctl deployment status [--json]. Keep this separate from the ClawHub product version and provenance.
  • For deliberate low-risk output or zone administration, use only the promoted soundclawctl outputs, soundclawctl zones, and explicit soundclawctl runtime reload [--json] command shapes listed in the local reference. Clarify the target and proposed mutation before acting.
  • For one on-demand Layer, use the promoted soundclawctl familiar run-layer or stop-layer command. For a bounded Scene, use the promoted soundclawctl scene apply/show/explain/stop boundary. Do not own execution state, retries, policy, or replay.
  • Guide source-preserving library ingest only through a promoted soundclawctl library ingest [--json] or public-bundle pi-kit flow after explicit operator approval. Do not browse arbitrary paths or edit the deployed library directly.
  • Use the optional private soundclaw-feedback capability only when it is present and the operator explicitly requests feedback lifecycle work. Never store feedback in this public skill.

General Behavior

  • Ask only for missing asset, output, Scene, Layer, volume, or topology facts needed for a safe deterministic command.
  • Prefer the smallest promoted read needed to resolve ambiguity.
  • Relay runtime-owned results and errors. Do not claim success from intent or marker checks alone.
  • Keep skill provenance, backend deployment identity, and runtime health as separate facts.

Out Of Scope

  • Backend installation, update, repair, rollback, or service manipulation
  • Copying, rsyncing, Git-syncing, or shadowing installed skills
  • Private Web UI routes, DOM state, runtime IPC, or host-file parsing
  • Invented runtime commands or local substitutes for unavailable capabilities
  • Hidden retries for mutating commands

Local References

  • Read {baseDir}/references/README.md for the operator job, guardrails, and intent boundary.
  • Read {baseDir}/references/examples.md for example requests and expected behavior.
  • Read {baseDir}/references/compatibility.md for runtime dependency and packaging notes.