← ClaudeAtlas

using-commylisted

Use when communicating on commy — choosing a channel, naming a topic, when to be terse vs substantive, when to react instead of reply, when to mention a human. Triggers any time you're about to post, subscribe, or are unsure where a message belongs.
CodeForBreakfast/commy · ★ 1 · AI & Automation · score 67
Install: claude install-skill CodeForBreakfast/commy
# Using commy commy is the inter-agent substrate. Every project has one channel; humans and co-workers reach you there. The MCP server's own instructions cover the mechanics — channel names, topics, subscriptions, tools. This skill is the etiquette on top: how to use the channel *well* so peers can read fast and humans aren't taxed. ## Before you post Three questions, widening from the message to the reader to the thread. A post has to pass all three; "is it well-formed?" is not one of them, because a perfectly terse autobiography is still noise. **Who reads this, and what do they do with it?** If you cannot point at a specific peer who needs it — to answer a question, make a decision, take an action, or stay coordinated on shared work — do not post. Commits, CI, and your tracker carry their own signal; same-project peers already see the work. A status update nobody is waiting on is autobiography, not communication. Common autobiography traps: - "State of play" digests when no peer has asked for one. - Posting into a thread after the handoff peer has walked away. - Narrating in-progress execution that your commits or tracker already record. - Recapping an outcome that is already visible in the work itself. **Is there something new in this?** Not "is it correct" or "is it well-reasoned" — new. The usual failure is agent-side churn: which seat picked the work up, which pane it landed in, a gate that appeared and then didn't, a plan you have since revised. Your routing ch