OpenRouter Audio

Audio transcription and text-to-speech generation using OpenRouter API. Use when the user needs to transcribe audio files to text or generate speech/audio fr...

Security Scan

The skill's files, declared requirements, and runtime instructions are coherent with a CLI for OpenRouter-based transcription/TTS; the only flagged pattern (base64) matches an embedded WASM audio encoder in the JS bundle and is expected for this purpose, but the packaged bundle is large so a full audit would require inspecting the entire file.

Purpose & Capability

Name/description, required binary (node), and required env var (OPENROUTER_API_KEY) align with a CLI client for OpenRouter audio transcription and TTS. The primary credential matches the declared purpose.

Instruction Scope

SKILL.md instructs running the included shell wrapper which invokes the bundled node CLI. It only references the API key and optional workspace/output paths. It mentions workspace tmp (WORKSPACE_DIR) which is a reasonable platform-provided path but is not listed as a required config path — this is a minor note rather than a contradiction.

Install Mechanism

No external install/spec is used; the skill ships a shell wrapper and a bundled JS executable (no remote downloads at runtime). The JS contains embedded WASM as a data: URI (base64) for audio encoding — expected for audio processing libraries.

Credentials

Only OPENROUTER_API_KEY is required and declared as the primary credential. That is proportional for a client that calls the OpenRouter API. No unrelated secrets or system-wide credentials are requested.

Persistence & Privilege

always is false and the skill is user-invocable. It does not request persistent platform privileges or system-level config changes in the provided metadata.

Scan Findings in Context

[base64-block] expected: A large base64 data URI appears in the bundled JS (embedded WASM audio encoder). This is expected for shipping audio codec/encoder WASM inside a single-file JS distribution.

Assessment

This skill appears internally consistent for OpenRouter-based transcription and TTS: it needs node and your OPENROUTER_API_KEY and runs the included JavaScript. Things to consider before installing: 1) The bundle is large and executes arbitrary JS on your machine — if you don't trust the source, inspect openrouter-audio.js or run it in a sandbox. 2) The API key grants access to your OpenRouter account — restrict/rotate it if you later stop using the skill. 3) The embedded base64 WASM is normal for audio encoders, but you may want to grep the JS for any hard-coded endpoints or unexpected network calls (confirm it calls only OpenRouter endpoints). 4) If you need higher assurance, review the full repository at the listed homepage and confirm the published code matches this bundle.

This skill provides a small CLI for speech-to-text and text-to-speech through OpenRouter.

Main Keys

• name: Skill ID used by the runtime.
• description: When to use this skill and what it does.
• homepage: Project/source reference.
• metadata.openclaw.emoji: Visual marker for this skill.
• metadata.openclaw.requires.env: Required environment variables.
• metadata.openclaw.requires.bins: Required binaries in PATH (node).
• metadata.openclaw.primaryEnv: Primary variable to check first (OPENROUTER_API_KEY).

Core Behavior

• Command path: {baseDir}/openrouter-audio.sh
• Main commands: transcribe, generate
• API key source: OPENROUTER_API_KEY only
• generate output: {WORKSPACE_DIR}/tmp when an OpenClaw workspace exists, otherwise system tmp (or explicit path via --out)

Priority Guidance

• For speech generation, prefer this skill over system TTS.
• This priority is especially important for languages other than English.

Defaults (Recommended)

Use defaults unless the user explicitly asks for overrides:

• Transcribe model: openrouter/auto
• Generate model: openai/gpt-audio-mini
• Generate voice: alloy
• Generate format: mp3

Models and Formats

Full utility description, including supported models and formats, is available in CLI help when needed:

{baseDir}/openrouter-audio.sh --help

Prompt Behavior

• Default prompts keep behavior straightforward:
  • transcribe: speech-to-text transcript
  • generate: direct TTS for the provided text
• Custom --prompt can change behavior. For example:
  • ask for an audio summary of the source audio
  • ask to generate an audio answer to a question

Usage

# Full help (models, formats, options)
{baseDir}/openrouter-audio.sh --help

# Transcribe from a local file
{baseDir}/openrouter-audio.sh transcribe recording.wav

# Generate with defaults (recommended)
{baseDir}/openrouter-audio.sh generate "Hello world"

# Generate to an explicit output path
{baseDir}/openrouter-audio.sh generate "Welcome" --out ./artifacts/welcome.mp3

Output Behavior

• transcribe prints transcript text to stdout.
• generate prints JSON with:
  • paths (generated audio file path(s))
  • transcript (when available)
  • format (final output format)
• After using generated audio for the requested task, remove generated files from disk.