Guide · Agent skill
How to use the Voice Note Ingest agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
Voice Note Ingest is an agent skill for the gbrain knowledge base. When you send a voice note or audio message, it tells your AI agent to keep your exact spoken words as a verbatim quote, store the original audio, and file the note in the right folder with links to any people or companies mentioned. This guide explains what the skill does, how to install it, how to trigger it, and what it needs before it will work.
Voice Note Ingestgarrytan/gbrain · MIT Download the skill Quick start
- 1Install gbrain and a storage backend. Confirm you have a gbrain knowledge base with S3, Supabase Storage or local storage configured. The skill cannot store audio without it.
- 2Install the skill. Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
- 3Check transcription. If your host does not supply transcripts, add a Groq Whisper or OpenAI Whisper API key. Split audio over 25MB with ffmpeg first.
- 4Send a voice note with a trigger phrase. Attach the audio and include a phrase such as "voice note ingest this" or "save this audio note".
- 5Check the routed page. Open the folder the agent chose and confirm the User's Words block matches what you said.
- 6Handle failures. If the audio was not stored, the skill files the note text and tells you the audio is missing. If a write conflicts, it re-reads and merges rather than rewriting your words.
What is an agent skill, and what does this one do?
An agent skill is a folder containing a SKILL.md file. The file gives an AI agent instructions for a specific task. When a request matches the skill's triggers, the agent follows those instructions.
This skill covers one job: ingesting a voice note into a gbrain knowledge base. It does not summarise the recording. The skill's core rule is that the user's exact words are the signal, so the agent keeps the transcript verbatim, including hesitations and filler words like "um" and "you know".
- Stores the original audio in the gbrain storage backend
- Transcribes the audio if the host does not supply a transcript
- Chooses a destination folder using a decision tree
- Writes a brain page with the verbatim transcript in a block-quoted "User's Words" section
- Adds timeline back-links to people or company pages it mentions
Install the Voice Note Ingest skill
Download the ZIP for this skill from this page. Then use one of the two options below.
For Claude Code, unzip the file into ~/.claude/skills/ to make it available for all your projects, or into a project's .claude/skills/ folder to limit it to that project.
In the Claude apps, upload the ZIP under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside your project folder
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Set up what the skill needs before you use it
The skill depends on gbrain. It needs a configured storage backend, which the source lists as S3, Supabase Storage or local storage. It also uses gbrain CLI commands such as gbrain storage and gbrain write-request. If you are not already running gbrain, this skill will not work on its own.
Transcription is optional. If the host agent already supplies a transcript, no transcription key is needed. If it does not, the skill suggests Groq Whisper or the OpenAI Whisper API, which require their own API keys. For audio over 25MB, the source says to split the file with ffmpeg first.
- gbrain knowledge base with a configured storage backend (S3, Supabase Storage or local)
- Optional: Groq Whisper or OpenAI Whisper API key, only if no transcript is provided
- Optional: ffmpeg, for splitting audio over 25MB
- gbrain CLI available for storage and write-request commands
How to trigger the skill with natural requests
You do not need a special command. Send the audio with a request that matches one of the skill's trigger phrases. The source lists these triggers: "voice note", "ingest this voice memo", "transcribe and file", "voice note ingest", "save this audio note" and "audio message".
The skill is designed for one note at a time. Each note is its own ingest cycle, and the source says there is no batching.
- "Voice note ingest this, it's about my plan for the studio."
- "Transcribe and file this audio message."
- "Save this audio note and link any people I mention."
- "Ingest this voice memo as a product idea."
How the agent decides where a note goes
The skill applies a decision tree and uses the first match. If a note fits several categories, it creates the primary page and cross-links the others.
Original ideas go to originals/. Observations about a named person go to people/, and about a company to companies/. Product or business ideas go to ideas/. Personal reflections are appended to personal/. Anything that does not fit goes to voice-notes/ as a dated catch-all file.
- originals/: your own novel thought, framework or connection
- concepts/: a framework or model someone else created that you reference
- people/ and companies/: new information about a person or company
- ideas/: something that could be built
- personal/: emotional, identity or reflective material
- voice-notes/: everything else, named YYYY-MM-DD-slug
Read the output you get back
Each brain page has a front-matter block with title, type, dates, tags and the storage path of the audio. The body contains a summary, the block-quoted "User's Words" section, a 🔊 Audio link, an Analysis section written by the agent, and a See Also list.
The source is firm that the transcript is not edited. The Analysis section is where the agent's interpretation goes.
- User's Words: the verbatim transcript, marked as not to be edited
- Analysis: the agent's interpretation, kept separate from the transcript
- Timeline: a dated entry for each voice note
What people use it for
- Capturing product ideas on a walk and filing them to ideas/ with the wording intact
- Recording a reflection that gets appended to a personal/ page
- Logging a conversation about a colleague to their people/ page with a timeline entry
- Keeping founder notes about a company in companies/ with back-links from that company's page
- Building source material for essays from raw, unpolished spoken thinking
Frequently asked questions
Is the Voice Note Ingest skill free?
The skill is published under the MIT licence, as stated on this page. Running it may still cost money, because the source mentions optional paid transcription APIs such as Groq Whisper or OpenAI Whisper, and any storage backend you choose.
What do I need to use the voice-note-ingest skill?
You need a gbrain knowledge base with a configured storage backend (S3, Supabase Storage or local). You also need gbrain CLI commands available. A transcription key is only needed if your host does not provide a transcript.
How do I install the voice note ingest skill in Claude?
Download the ZIP from this page. In Claude Code, unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder. In the Claude apps, upload it under Settings → Capabilities → Skills.
Does this skill summarise my voice notes?
No. The skill forbids paraphrasing and summarising the transcript, so the output keeps your exact words, including filler words. Expect long verbatim text rather than short summaries.
Is it safe to use with sensitive recordings?
Be cautious. The skill uploads audio to a storage backend and may call external transcription APIs. Do not use it for sensitive recordings you do not want stored or sent to those services.
What if the agent cannot store my audio?
The source says the agent files the note text it has and tells you the audio was not stored. Check that your gbrain storage backend is configured, since the skill relies on gbrain storage commands.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use Voice Note Ingest if you already run gbrain and want voice memos filed consistently, with your exact wording preserved and linked to people and companies. Pick something else if you only want a short summary, if you do not run gbrain, or if you are uneasy about audio being stored or sent to a transcription service.
Download the skill