Skip to main content

followup-sequence

Generate and schedule a 4-touch follow-up cadence where each touch adds a new reason — and automatically stop the moment the prospect replies.

Datos de origen

Repositorio
ryanfrigo/clawmart
Última actividad en el origen
3 de julio de 2026 a las 06:45
Idioma detectado de SKILL.md
inglés
Estrellas
3
Forks
0

Opciones de instalación

De forma predeterminada está seleccionado el prompt que primero revisa el origen. Puedes cambiar a un comando directo o descargar una copia local.

Revisa los archivos de origen

Lee SKILL.md y los archivos complementarios que muestra SkillsMP antes de decidir si quieres instalarlo.

Mostrando SKILL.md

SKILL.md
Instrucciones de origen · Vista previa de solo lectura
name
followup-sequence
description
Generate and schedule a 4-touch follow-up cadence where each touch adds a new reason — and automatically stop the moment the prospect replies.
metadata
{"openclaw":{"emoji":"🔁"}}
# Follow-up Sequence Most replies come after the first email, not on it. This skill builds a disciplined 4-touch cadence where every follow-up adds a *new* angle instead of guilt-tripping — and, critically, stops the instant the prospect replies or opts out. ## When to use (trigger phrases) Use this skill when the user says: - "set up follow-ups for <prospect>" - "queue the cadence" / "sequence this lead" - "what's my next follow-up to <name>?" - "schedule 4 touches for <company>" - (Handoff) right after `cold-open` drafts touch one. ## How it works 1. **Anchor on touch one.** Take the `cold-open` message as touch 1 (day 0). If it hasn't been sent, note that the cadence starts when it is. 2. **Design four touches, each with a distinct angle.** Never "just bumping this to the top of your inbox." Every touch must justify its own existence: - **Touch 1 — day 0:** the cold open (the observation + hypothesis + soft ask). - **Touch 2 — day +2 or +3:** a *new* angle. A different signal, a specific resource ("we wrote up how a team your size handled this"), or a sharper version of the value hypothesis. Reply in the same email thread. - **Touch 3 — day +5 to +7:** proof or a pattern. A concrete example of the outcome (no fabricated logos/case studies — use real ones or a specific mechanism), or a one-line "here's what week one looks like." - **Touch 4 — day +10 to +14:** the polite breakup / permission to close. "Sounds like this isn't a now problem — I'll close the loop unless you'd rather I check back next quarter." Breakups often get the highest reply rate because they hand back control. 3. **Respect timing rules.** Send during the prospect's business hours (use their timezone from the research card), skip weekends and obvious holidays, and never send two touches on the same day. Space follow-ups by working days, not calendar days. 4. **Thread correctly.** For email, all follow-ups reply within the same thread (same subject, `Re:`) so the prospect has context. For LinkedIn/DM, keep each touch even shorter. 5. **Register the stop conditions — this is the most important step.** The cadence must cancel all remaining touches on any of: - A reply from the prospect (hand off to `reply-triage`). - A hard bounce or "no longer at this company." - An unsubscribe / "remove me" / "not interested" (also suppress — see Notes). A bot that keeps sending after someone replies is the fastest way to burn a domain and a reputation. 6. **Persist the schedule.** Write each scheduled touch to `pipeline-log` with its send date and status (`scheduled` / `sent` / `canceled`). The daily run checks what's due today; that's how "scheduling" actually works without a dedicated sequencer. 7. **Cap the cadence.** Four touches, then stop and mark the prospect `nurture` (recheck in a quarter). Do not invent a fifth, sixth, seventh touch — that's harassment, not persistence. ## Output The full sequence with dates, angles, and the stop rule. Example: ``` CADENCE — Dana Ruiz @ Acme (starts on send of touch 1) Touch 1 Day 0 [cold-open] Angle: hiring signal → ramp gap Touch 2 Day +3 Tue 9:15a PT Angle: resource — "AE ramp w/o an SDR" Touch 3 Day +7 Mon 8:40a PT Angle: mechanism — what week one looks like Touch 4 Day +12 Thu 10:00a PT Angle: breakup — permission to close STOP on: any reply, bounce, or opt-out → cancel remaining, notify user. State written to pipeline-log (status: scheduled). Draft of Touch 2: (in-thread reply, no new subject) One more thought, Dana — the piece that usually bites new AEs isn't the calls, it's the 40 minutes of research before each one... ``` ## Notes - **Scheduling depends on your setup.** OpenClaw doesn't ship a native drip sequencer. This skill schedules by writing dated tasks to `pipeline-log`; a daily agent run (or a cron you configure) sends what's due. If you've connected a real sequencer or CRM, prefer that. - **Stop-on-reply requires inbox access.** The stop condition only fires if the agent can see replies and bounces. Configure the same inbox the cold opens send from, and let `reply-triage` watch it. - **Honor opt-outs immediately and permanently.** "Remove me," "unsubscribe," or "stop" must cancel the cadence *and* suppress the contact from all future outreach — same day. This is both basic decency and, in most jurisdictions, the law (CAN-SPAM, GDPR/PECR). - **No fake urgency or fake proof.** Don't fabricate "only 2 spots left" or case studies that don't exist. The breakup email works because it's honest, not because it's a trick. - **Volume discipline.** Spreading four touches per prospect over two weeks keeps daily send volume and spam risk manageable. Don't compress the cadence to hit someone five times in three days.
Ver en GitHub