agentleFS
TrustPagerTrustPager/Business_Operating_System/skills/draft-reply/SKILL.md

draft-reply

Skill8 starsChanged 5 months agoMITView on GitHub

Draft a reply to an email or SMS you received — pulls the context, writes the response in your tone, shows it for your approval.

6 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 8 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If the user pastes a message directly (not from the JSON), use that paste as the source-of-truth and skip the fetch.

    line 40
  2. Distinctive:

    Before drafting, identify in the inbound:

    line 44
  3. Distinctive:

    If unclear ("I'm not sure what they're actually asking"), say so to the user and ask for clarification BEFORE drafting:

    line 49
  4. Distinctive:

    Acknowledge first, then answer: If they raised a concern, name the concern before responding to it.

    line 54
  5. Distinctive:

    If the inbound has multiple questions, address them in bullet order.

    line 60
  6. Distinctive:

    Wait for yes/no

    line 66
  7. Distinctive:

    When you don't know, say so: If they asked "what's the price?" and you don't know, the reply says "Let me confirm and come back to you …

    line 73
  8. Distinctive:

    Objections and 'no's have doctrine: If the inbound is a price objection or a decline, read business-method.md §9.4 and §12.5 before …

    line 74

“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/draft-reply

Replies are higher-stakes than fresh emails — the context already exists, and getting the tone wrong is more obvious. This skill drafts the reply against the actual message that was received, not against a guess.

Step 1 — Find the message being replied to

First, run:

python ~/.claude/bos-run.py draft-reply --hours 48

This returns every inbound email + SMS in the window with no reply, ranked (open-opportunity senders first, then by recency). If the user already named someone, filter the list and pick that one. If not, offer the top 3-5 as a numbered list:

"These are the unanswered messages from the last 48h, top first. Reply to #1? Or pick another?"

If the user pastes a message directly (not from the JSON), use that paste as the source-of-truth and skip the fetch.

Step 2 — Read the inbound carefully

Before drafting, identify in the inbound:

  • What they're asking — a question, a request, an objection, a confirmation, an emotional venting?
  • The emotional register — formal? casual? frustrated? excited?
  • Specific details — names, dates, dollar amounts, products mentioned — these must appear in the reply.

If unclear ("I'm not sure what they're actually asking"), say so to the user and ask for clarification BEFORE drafting:

"I'm reading this as a request for [X] — is that right, or are they actually asking [Y]?"

Step 3 — Draft the reply

  • Acknowledge first, then answer. If they raised a concern, name the concern before responding to it. ("Totally hear you on the timing — here's what we can do…")
  • Match their tone. They were casual? Be casual back. They were formal? Don't get matey.
  • Answer the question they asked, not a question you wish they'd asked. If they asked "is it $5k or $7k?" the reply leads with the number, not with how the pricing works.
  • One next step. Like /send-email, end with one clear ask.
  • No fake enthusiasm. "Great question!" / "Love this!" / "So excited to hear from you!" — banned.

If the inbound has multiple questions, address them in bullet order. Keep the same structure they used so the reply maps obviously to their email.

Step 4 — Show + send

Same approval flow as /send-email:

  • Show the proposed reply (subject = same as inbound prepended with "Re:", or just the SMS body)
  • Wait for yes/no
  • On yes: mcp__trustpager__reply_to_email (for email) or mcp__trustpager__send_sms (for SMS)
  • On no: ask what to change. Common edits: "more direct", "less formal", "shorter", "include the dollar number".

Important behaviours

  • Content guardrails. Customer-facing copy uses no em dashes, invents no facts, quotes, or numbers, and names no third-party vendor. Write it in the owner's brand voice; the framing and the marketing psychology are the owner's choice. The rules are in knowledge/content-rules.md. The register for these messages is knowledge/communication-voice.md: plain, warm, short.
  • When you don't know, say so. If they asked "what's the price?" and you don't know, the reply says "Let me confirm and come back to you today."
  • Objections and 'no's have doctrine. If the inbound is a price objection or a decline, read business-method.md §9.4 and §12.5 before drafting: never offer the same thing cheaper; a genuine alternative changes scope or payment shape. Restate their concern in the reply before answering it.
  • Threading matters. Email replies must use reply_to_email (preserves the thread), not send_email to the same address (breaks threading).
  • CCs in the reply. Preserve the inbound's CC list unless the user removes one. Don't silently drop CCs.
  • The user is the final filter. Even after edits, every send is approved.

Output shape

"Replied to {sender} on '{subject}' — logged on {opp_name if linked}."

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.