Habit Agent Prompts That Work: Nudges, Snoozes, and Resets

Most agent prompts fail because they’re written like notifications:

"Time to do X."

But an agent reminder is a conversation. The prompt needs to make cooperation easy — offer exits (done/snooze/skip), respect context (windows, focus time), and stay short enough to read in a busy channel.

This post is a micro-library of prompts you can reuse for habit agents and todo agents — especially in OpenClaw-style workflows where agents show up across ChatGPT, Telegram and Slack.

How to use this library

1. Pick the situation — nudge, completion, snooze, skip, reset, or last-chance follow-up.
2. Pick the channel — Telegram (short, personal), Slack (lighter, work-visible), ChatGPT (longer, reflective).
3. Swap placeholders — {habit}, {time_window_end}, {next_step}, {tomorrow} (keep names consistent with your activity titles).
4. Ship one habit at a time — tune wording after a week of real replies; copy-paste beats crafting novel copy every day.

Rules everywhere: one ask per message, one default next step (e.g. snooze 20m), one skip exit without guilt, and name the activity (“Morning stretch window”, not “Reminder”).

If you want the OpenClaw starting point:

• OpenClaw Habit Agent

Copy-paste starters (Telegram vs Slack)

Use these as first lines; adapt placeholders to your rituals.

Telegram (DM, mobile):

• Nudge: Window’s open — {habit}. Now, snooze 20m, or skip today?
• Snooze ack: Got it — back in 20m. I’ll go quiet after this window.
• Skip: Skipped. Same plan tomorrow or shift the window this week?

Slack (thread or small ritual channel):

• Nudge: Ritual check: {habit} — done / snooze 20m / skip today
• Snooze ack: Snoozed 20m — last ping before window closes unless you reply.
• Skip: Logged skip — no drama. Want the same cadence next week?

For shared channels, avoid @channel unless the culture expects it; use a dedicated #rituals channel when you can (Slack Routine Bot vs Agent).

Nudges (first reminder)

• "Window’s open. Want to do {habit} now, or shift it later?"
• "Quick check: do you want {habit} now, or should I remind you closer to {time_window_end}?"
• "Two-minute version today, or full version later?"

Completions (make ‘done’ easy)

• "Nice. Marking {habit} done."
• "Done. Want to keep momentum with {next_step}, or pause here?"

Snoozes (default + flexible)

• "Got it. Snoozing {habit} for 20 minutes."
• "Snooze for 20m, or pick a time: 30m / 1h / later today?"
• "If you’re in a focus block, say ‘quiet mode’ and I’ll hold reminders until it ends."

Skips (data, not shame)

• "No problem—skipping today. Want to keep tomorrow the same?"
• "Skipped. If this is happening often, we can adjust the window."

Resets (when a streak breaks)

• "Streaks break. Want to treat tomorrow as a clean restart?"
• "If this habit isn’t fitting your week, should we shrink it (2-minute version) or move the window?"

Gentle follow-ups (avoid spam loops)

• "Last check in this window: do you want to do {habit}, or skip today?"
• "I’ll go quiet after this. If you want to reschedule, say ‘move to {tomorrow}’."

Example conversations (short)

Habit reminder

Buffy: "Morning startup window is open. Water + stretch now, or plan first?"
You: "Snooze 20."
Buffy: "Done. I’ll check back in 20m."

Task follow-up

Buffy: "Your report is due tomorrow. Want a 15-minute start now, or schedule a focus block?"
You: "Schedule a focus block."
Buffy: "Got it—what time window works best?"

Channel variants: matching tone to surface

The same situation — a nudge, a snooze, a skip — calls for different phrasing depending on where it lands.

Why the tone varies

Telegram is personal and interruptive — the user is on their phone, probably mid-activity. Every word counts. Keep it to one sentence and a binary choice.

Slack competes with real work messages in a shared channel. Prompts should feel lighter, less urgent, and slightly more "team-awareness" in tone — even for personal habits. Never guilt.

ChatGPT is already a reflective conversation. The user is there intentionally. Prompts can be longer, explain context, and invite real discussion about changing the habit. This is also where resets and redesigns happen best.

Channel routing rule of thumb

• Telegram: execute and log (short replies, mobile-first)
• Slack: coordinate and close loops (work-context habits and team routines)
• ChatGPT: design and reflect (set up habits, review patterns, redesign windows)

For more on channel selection, see: Multi-Channel Habit Tracking Across ChatGPT, Telegram and Slack

Where this fits in Buffy’s design

These prompts work best when they’re connected to a behavior engine that understands:

• activities (habit/task/routine)
• time windows + context
• event history + memory

Related posts:

• OpenClaw Agent Reminder UX
• Designing Conversational Reminders That Don't Annoy You

Where to go next

• Next step: set up one habit with a time window, then use these prompts for nudges/snoozes/skips: How to Get Started With Buffy Agent in 5 Minutes

Further reading

• OpenClaw Habit Agent
• ChatGPT for Planning, Telegram for Execution
• OpenClaw Habit Agent Memory: Why Chat Context Isn’t Enough