Speech to Text Transcription
ReviewAudited by ClawScan on May 1, 2026.
Overview
This transcription skill is purpose-aligned, but users should be aware that cloud transcription sends audio to third-party providers and the skill keeps local preference memory.
Before installing, decide whether your recordings are sensitive. Prefer local Whisper for private audio, confirm before using cloud transcription, keep API keys in environment variables, and periodically review the local ~/speech-to-text-transcription/ folder.
Findings (5)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
If you choose a cloud provider, your audio may contain private conversations, meetings, or other sensitive content that leaves your device.
The skill clearly discloses that audio may be sent to external transcription providers when cloud APIs are used.
Data that leaves your machine (if using APIs): Audio file sent to chosen provider (OpenAI, AssemblyAI, Deepgram)
Use local Whisper for sensitive audio, and only use cloud providers after confirming you trust the provider and its retention/privacy terms.
Using these keys can consume paid API quota and grants the skill access to submit audio to the selected provider.
The skill may use cloud transcription account credentials, but the need is disclosed and aligned with the provider integrations.
Optional API keys (only if using cloud providers): OPENAI_API_KEY, ASSEMBLYAI_API_KEY, DEEPGRAM_API_KEY
Use environment variables as documented, avoid hardcoding keys in files, and prefer limited-scope or dedicated API keys where available.
Preference memory may reveal patterns such as common use cases, language, or output choices, even if full transcripts are not saved automatically.
The skill maintains persistent local memory and may learn preferences from transcription activity, while separately stating transcripts are saved only on request.
Gather context from each transcription ... Save transcripts only when asked
Review or delete ~/speech-to-text-transcription/memory.md if you do not want transcription preferences retained.
Running these commands can install software and create or modify local media/transcript files.
The skill documents local package installation and shell commands for transcription and media conversion; these are central to the skill's purpose and appear user-directed.
pip install openai-whisper ... ffmpeg -i video.mp4 -vn -acodec pcm_s16le -ar 16000 -ac 1 audio.wav
Approve package installs and file-processing commands deliberately, and check input/output paths before running them.
A user might overestimate privacy guarantees for audio sent to third-party transcription services.
The skill also warns users to trust cloud providers, but this retention statement should not be read as a guarantee about each provider's independent data-retention policy.
This skill does NOT: ... Retain files on external servers after processing
Check the selected provider's retention policy before uploading sensitive recordings.