reply-triage
Skill0 starsChanged 50 days agoMITView on GitHub
Use when the user wants to go through their Warmerly inbox, find replies that need an answer, sort interested leads from autoresponders and unsubscribes, or draft and send replies to campaign responses. Requires the Warmerly MCP server.
In this repository3 other files it gives its agents.
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 14
You are writing to real people as the user: Every reply goes out from their mailbox under their name.
- Very specific: line 53
Use categorize_message with lead, not_relevant or spam to file each conversation, so the user's inbox matches your sort.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Reply triage
Go through the conversations that are waiting on the user, sort them honestly, and answer the ones worth answering. Speed matters for a warm reply; accuracy matters more. One wrong reply to someone who asked to be left alone does more harm than ten slow ones.
Guardrails (read first)
- You are writing to real people as the user. Every reply goes out from their mailbox under
their name. Unless the user has explicitly told you to send on your own, show each reply and
get a yes before
send_reply. Drafting is always fine; sending is the step that needs consent. - Never auto-reply to an unsubscribe request, a complaint, an angry or threatening message, or a legal notice. Suppress the address instead (below) and tell the user. A reply to "stop emailing me" is itself the next complaint.
- Respect plan limits. AI drafting spends AI credits. If a tool refuses on a limit, say so and point to https://app.warmerly.com/settings/usage.
- Never invent facts in a reply: prices, availability, case studies, meeting times. If the answer needs something you do not know, ask the user or leave a clearly marked gap.
Step 1: What is waiting
list_conversations with filter set to needs_reply (the newest message is inbound and the
contact is a lead). Use unread for "anything new", and q to search for one person or company.
Up to 50 per call; say how many there are in total.
Step 2: Read before judging
get_conversation for each one. It returns the timeline (latest 50 messages), any saved draft,
and which mailbox a reply should send from. Reading through the tools does not mark anything as
read for the user.
Step 3: Sort each one
| Kind | How to spot it | What to do |
| Interested | asks a question, wants a call, asks for pricing or details | Draft a reply. Top of the list for the user. |
| Referral | "talk to Jane, she handles this" | Draft a short thank-you; suggest adding Jane as a lead (do not email her without the user's yes). |
| Not now | "maybe next quarter", "not a priority" | Short, gracious reply; note the timing for the user. |
| Not interested | a polite no | Usually no reply needed, or one line of thanks. Categorize not_relevant. |
| Unsubscribe / angry / complaint | "remove me", "stop", "unsubscribe", "how did you get my address", anger, legal threats | Do not reply. suppress_emails with their address, categorize not_relevant, tell the user. |
| Autoresponder | out of office, "I'm away until", ticket numbers, "this mailbox is not monitored", "no longer with the company" | Not a reply. Do not answer and do not count it as a response. If it names a replacement contact, mention it. If the person has left, suggest suppressing the old address. |
| Bounce / delivery failure | mailer-daemon, "delivery failed", "address not found" | Not a reply. Suppress the address. |
| Spam | unrelated selling, phishing | Categorize spam. |
Most campaign "replies" are autoresponders. Say so when you report numbers, so the user does not read an out-of-office wave as interest.
Use categorize_message with lead, not_relevant or spam to file each conversation, so the
user's inbox matches your sort.
Step 4: Draft
draft_reply generates a draft from the conversation. Then edit it to these rules before showing
the user:
- Answer the question they asked, first. Two to five sentences.
- Plain text, no em dashes, every sentence starting with a capital letter.
- One clear next step (a time to talk, a link, one question).
- Match their tone; if they wrote two lines, do not send six.
- Nothing that was not in the conversation or given to you by the user.
Step 5: Send
With the user's yes (or explicit autonomy), send_reply with the conversation's contact and the
final body. Send from the mailbox get_conversation named unless the user says otherwise; it keeps
the thread on the address the lead already knows.
Step 6: Report
Finish with a short summary: how many conversations, how many were real replies, who is interested (name, company, what they want), what you sent, what you suppressed, and what needs the user's own answer.
When a tool is missing
draft_reply, send_reply, categorize_message and suppress_emails may not be available on
the user's connection yet. If one is not in your tool list, do not pretend to call it. Write the
draft in the chat for the user to paste, and send them to https://app.warmerly.com/inbox to send
it and file the conversation, and to https://app.warmerly.com/settings/suppression to add
addresses to the suppression list.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.