Meeting notetaker

Skill theDakshJaitly/mex-call/skills/meeting-notetaker

How and when to use mex-call — a bot that joins a Google Meet, captures the conversation as structured agent-readable memory in the repo, and acts on the "Mex, …" wake phrase. Load this when the user mentions joining/recording a meeting or call, taking meeting notes, capturing decisions from a call, a standup/user-research/design review, or asks what mex-call can do. This skill is guidance only — it never joins a call by itself; launching is always user-triggered via /mex-call:call.From its SKILL.md

Install

npx -y skills add theDakshJaitly/mex-call --skill meeting-notetaker

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.8 KB, as published. Nobody here has run it

Using mex-call

mex-call sends a notetaker bot into a Google Meet. It listens, continuously writes bounded structured memory under .mex/meetings/ in the current repo (rolling summary, decisions, action items, open questions, participants), and when someone says "Mex, …" out loud it answers in the meeting chat, records a decision, or takes a repo action ("Mex, create an issue for that" → a real GitHub issue / doc edit / PR draft, grounded in the call). It does not speak. It is built on mex but never requires it — in a repo without a mex scaffold it runs standalone and nudges; in one with a scaffold it also reads .mex/context and .mex/patterns.

When to suggest it

If the user is about to join, or is on, a meeting they want captured into the repo — standups, user-research calls, design/architecture reviews, client calls, or even solo rubber-ducking — suggest launching mex-call. The value is that the conversation stops being ephemeral and becomes durable context attached to the code.

Do not launch it yourself. Joining creates a real (billed) bot in a live call — that is a side effect the user must trigger. Point them to the launcher instead.

How to launch

/mex-call:call <google-meet-link> — this launches the runtime in the background and turns the session into a live dashboard. The user must admit "Mex (notetaker)" when it knocks. (Requires RECALL_API_KEY in the repo's .env or ~/.mex-call.env.)

Driving it during a call

The dashboard (.mex/meetings/live/dashboard.md) shows status, participants, rolling summary, decisions/actions/questions, and a live activity feed. The plugin's background monitor streams each new activity line into the session.
mex-call watch — a continuous terminal dashboard.
mex-call leave — make the bot leave and archive the call to .mex/meetings/<date>-<name>/.
The bot only acts on the explicit "Mex, …" wake phrase; otherwise it just listens and writes memory. Output is chat + files — never voice.

After the call

The archived folder holds the final summary, decisions, action items, open questions, and full transcript — durable, agent-readable context for future work in this repo. If the user asks "why did we decide X?", look there.