From founder-os
Draft an email in the founder's voice. Trigger on "write an email", "draft a reply", "send a message", "follow up with", "reach out to", "respond to this email", "compose an email", or any variation. Also fires when the user pastes email content and asks for a response, or when they describe a situation that clearly needs an email. Covers client emails, partnership outreach, investor updates, team communications, cold outreach, and personal correspondence. Reads `core/voice-profile.yml`.
How this skill is triggered — by the user, by Claude, or both
Slash command
/founder-os:email-drafterThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Runs on: reasoning - reads your files and reasons; any capable agent can run this.
Runs on: reasoning - reads your files and reasons; any capable agent can run this.
Apply the voice and writing rules from core/identity.md, rules/writing-style.md, and core/voice-profile.yml (via the your-voice skill). Emails are conversations, not content pieces. The anti-AI self-check is relaxed here, but the voice rules still apply.
Apply the routing rules in skills/your-voice/SKILL.md "Voice routing - operator or brand?" section. Default to operator voice for personal emails. If the user names a brand they run (e.g. "draft an email from <brand> to a customer"), use brand voice from brands/<slug>/voice.yml.
If using operator voice, run: python scripts/check-voice-ready.py
If using brand voice, run: python scripts/check-brand-voice-ready.py --brand <slug>
If neither operator nor brand voice is explicitly named, default to operator voice and run python scripts/check-voice-ready.py. Stop on exit code 1.
If exit code is 1, read the output line and surface it to the user verbatim. Do not produce any draft. Stop.
If the user explicitly chooses to proceed with defaults after seeing that message, draft the email using the universal anti-AI baseline from your-voice and clearly label that the voice profile was not applied. Do not pretend the email is voice-coupled.
Then read the chosen voice profile so the rest of this skill can apply it.
After producing a draft and before returning it, run the anti-examples filter:
anti_examples.pairs block in core/voice-profile.yml.bad: pattern (literal substrings, structural markers like negation-contrast, or rule-of-three lists).good: pattern as the model and the rule: line as the constraint.aesthetic_crimes phrase or a red_flags pattern.Do not surface this filter to the user as a separate step. The user sees only the cleaned draft.
Before drafting, read brain/.snapshot.md if it exists. Use the open-flags block to avoid topics that contradict current operator stance. Use the must-do block to lean the draft toward what the operator is actively working on. Use the voice and brand blocks (if present) to set tone. If brain/.snapshot.md does not exist, proceed without it - the snapshot is optional context, not a hard prerequisite.
Read stack.json. The email_platform field is one of gmail, outlook, apple_mail, or null.
gmail: if a Gmail MCP is connected, offer to read the inbound thread for context. Otherwise ask the user to paste the prior message.outlook: same pattern via Outlook MCP. Do not assume Gmail.apple_mail or null: prompt the user to paste the prior message. No MCP integration available.Do not hardcode "Gmail" in instructions to the user. Mirror their actual platform.
Lead with the point. No warm-up paragraphs. No "I hope this email finds you well." No "I wanted to reach out regarding." Just say why you're writing in the first sentence.
One email, one ask. If you need three things, make the most important one the clear ask. The others can be mentioned but don't bury multiple action items in prose.
Write short. Most emails should be 3-8 sentences. If it's longer than a screen, it should probably be a meeting or a doc.
Sign off simply. Use the founder's first name from core/identity.md. For first-contact or formal situations, use a full block:
[Founder Name]
[Title] - [Company]
No inspirational quotes in signatures. No "sent from my iPhone" disclaimers.
If the user gives you a vague "write an email to X", ask:
If the user gives you enough context, just write it. Don't over-ask.
Bad (AI email): "I hope this email finds you well. I wanted to reach out to discuss the potential for collaboration between our organizations. I believe there could be significant synergies..."
Good: "Ahmed - saw your post about the ops bottleneck at scale. We solved something similar for a 40-person services company in Abu Dhabi last quarter. Worth a 15-minute call? I can share what worked."
Bad (AI follow-up): "I hope you're doing well. I'm following up on our previous conversation regarding the proposal I sent over. Please let me know if you have any questions or concerns."
Good: "Quick follow-up on the proposal from Tuesday. Two things changed since we spoke - updated the timeline on the systems audit and added the team structure review you mentioned. Let me know if you want to walk through it."
npx claudepluginhub arcassystems/founderos --plugin founder-osProvides UI/UX resources: 50+ styles, color palettes, font pairings, guidelines, charts for web/mobile across React, Next.js, Vue, Svelte, Tailwind, React Native, Flutter. Aids planning, building, reviewing interfaces.
Fetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.