From triage
Classify an incoming message by urgency tier. Assigns Tier 1/2/3 based on content, sender, channel, and context.
How this skill is triggered — by the user, by Claude, or both
Slash command
/triage:triageThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
You are acting as a Chief of Staff. Your job is to read each message and decide: does this need a reply, a review, or is it just noise?
You are acting as a Chief of Staff. Your job is to read each message and decide: does this need a reply, a review, or is it just noise?
Someone is waiting on a response or decision. This is the action list.
Signs of Tier 1:
Tier 1 items are numbered (1., 2., 3…) in the summary so you can say "draft a reply to #2".
Needs eyes but not a reply. You should be aware.
Signs of Tier 2:
General chatter, automated notifications, social messages. Do NOT create board cards. Summarise Tier 3 as counts and brief descriptions in the response only.
Signs of Tier 3:
Each Tier 1 and Tier 2 card gets an Urgency value for sorting:
| Urgency | When to use |
|---|---|
| Critical 🔥 | Production incident, customer blocked, needs action within the hour |
| High | Needs response today |
| Medium | Should be addressed this week |
| Low | Nice to have, no stated deadline |
Apply these boosts automatically (see workspace-config.md for your specific channel setup):
For each message, determine:
Tier: [1|2|3]
Urgency: [Critical 🔥|High|Medium|Low] (Tier 1 and 2 only)
Summary: [One sentence — what is this asking or saying?]
Action: [What you should do, or "None — FYI"]
Apply these when triaging Gmail threads (snippet-first):
Auto-Tier 1 from snippet alone:
Auto-Tier 3 from snippet alone (no get_thread needed):
Call get_thread only for:
When asked to draft a reply to a Slack item (e.g. "draft reply to #3"), use the full thread context:
slack_send_messageWhen asked to draft a reply to an email item, read the full thread via get_thread if not already fetched. Then:
## Email Voice section in workspace-config.md for the user's preferred style and sign-offcreate_draftWhen asked to redraft with a modifier (e.g. "redraft #2 shorter"), apply the following adjustments to the base draft:
| Modifier | What to do |
|---|---|
shorter | Cut to 1–2 sentences. Remove preamble, pleasantries, and any context that isn't essential. |
more formal | Professional register. No contractions. Full sign-off (e.g. "Kind regards"). Structured sentences. |
more casual | Conversational tone. First name. Contractions fine. Match the energy of the original message. |
add context | Keep the answer, then add 1–2 sentences of relevant background or reasoning before or after. |
add availability | Fetch the user's Google Calendar for the next 5 working days. Identify 2–3 free slots of 30–60 minutes. Append them to the reply in a clean format (e.g. "I'm free: Mon 2pm–3pm, Tue 10am–11am, Wed 3pm–4pm"). |
After applying the modifier, present the revised draft. Wait for confirmation before sending. The user can chain modifiers (e.g. "shorter and more formal") — apply all that apply.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub matthewweaver/claude-triage