Voice Command to Skill

Overview

voice_command_to_skill is the voice-to-action bridge of the LabOS anywhere-lab stack. It takes natural language speech — transcribed by an ASR engine (Whisper, Azure Speech, Google Cloud) — and maps it to a specific LabClaw skill call with filled parameters. Commands like "check if I added the enzyme", "what's the next step?", "export my data to Excel", or "did I miss any steps?" are parsed into intent, matched to skills (protocol_video_matching, detect_common_wetlab_errors, extract_experiment_data_from_video, etc.), and executed with context-aware parameters. The skill provides prompt templates, parameter extraction logic, and fallback handling so that voice-driven lab workflows remain robust under noisy conditions, ambiguous phrasing, or partial context.

When to Use This Skill

Use this skill when any of the following conditions are present:

• Hands-free lab operation: A researcher is wearing XR glasses or has gloves on and cannot type or tap; they must control the system by voice — "run compliance check", "show me the growth curve", "pause protocol".
• Anywhere-lab / remote supervision: A PI or remote expert monitors a lab via video and issues voice commands to trigger analysis, generate reports, or request status — "extract the OD values from the last hour", "generate the Methods section".
• Training and onboarding: A trainee asks questions by voice — "what do I do next?", "did I do that right?", "explain step 5" — and the system routes to the appropriate skill for response.
• Post-experiment voice recap: After an experiment, the researcher speaks a summary request — "give me a report of what we did" or "check for any errors in the recording" — and the skill invokes report generation or error detection.
• Multi-modal AR interaction: Voice complements gaze, gesture, or touch in an XR lab interface; the skill resolves voice intent and coordinates with other input modalities.
• Accessibility: Researchers with mobility limitations rely on voice as the primary control channel for lab software and analysis pipelines.
• Rapid iteration: During protocol development, the researcher iterates by voice — "try that again with 50 microliters", "skip to step 8" — without breaking flow to use a keyboard.
• Batch command chaining: A single voice command triggers a multi-skill pipeline — "analyze the video and export to Excel" → analyze_lab_video_cell_behavior + export_experiment_data_to_excel.