3.1k Downloads
Overview
Lightweight, zero-LLM one-shot reminders (under 48 hours) scheduled via a nohup-based shell script that sleeps until fire time and then sends a message through `openclaw message send` to a specified channel/target.
Key Advantages
1.Zero LLM usage at delivery time — all wording is composed once at creation, minimizing latency, cost, and unexpected tone changes.
2.Simple, transparent CLI interface (`nohup-reminder.sh add|list|remove`) that fits well into OpenClaw/workspace tooling and automation.
3.Supports both relative and absolute times with optional timezone handling, covering most short-term reminder needs.
4.Multi-channel delivery support (Telegram default, optional WhatsApp/Discord/Signal/iMessage) via `--channel` and `--target`.
5.Strong, explicit guardrail to avoid misuse for >48h reminders by deferring to calendar tooling instead, reducing reliability risks for long-horizon tasks.`,`Encourages human, friendly reminder text ("
Use Cases
- Short-term personal reminders for tasks within the next 48 hours (e.g., call someone in 2 hours, grab laundry in 20 minutes, check the oven in 30 minutes).
- Lightweight follow-up nudges during ongoing workflows where a full calendar event would be overkill (e.g., "ping me in 45 minutes if the deployment isn’t done").
- Quick cross-channel pings (Telegram by default, or WhatsApp/Discord/etc.) for time-sensitive checks or errands.
- Temporary reminders during focused work sessions (Pomodoro-style breaks, stretch reminders, time-boxed tasks).
- Short-lived operational reminders for on-call or devops tasks (e.g., "check logs in 1h", "verify backup in 90m").
Evaluation Scores
7.8
/ 10
Reliability
7.0
Functionality
7.5
Usability
7.5
Safety
8.0
Performance
9.0
Compatibility
8.5
Based on 1 evaluation · Latest: 3/19/2026
Download Trend
Loading...
Evaluation History (1)
7.8/103/19/2026▼
OS: darwin-arm64LLM: z-ai/glm-4.5-air
**Quick judgment**: A strong, minimalistic skill for short-horizon (<48h) reminders. It integrates cleanly via a shell script, avoids LLM usage at fire time, and produces friendly, human-feeling reminder messages. Best suited for time-sensitive, near-term nudges rather than full task management.
**Key strengths**
- Zero-LLM delivery: once the reminder text is composed, no further model calls are needed, making delivery cheap, fast, and predictable.
- Clear CLI model (`add`, `list`, `remove`) with good examples and default `baseDir` fallback.
- Flexible timing: supports relative durations and absolute timestamps with IANA timezone support.
- Multi-channel messaging via `--channel` and `--target`, with guidance to use `TOOLS.md` and `session_status` to resolve delivery targets.
- Strong UX guidance for natural, context-resilient reminder text that reads like a friend, not a system.
**Main risks / limitations**
- Hard 2-day conceptual boundary: the skill is explicitly not for reminders 2+ days out; the agent must reliably defer those to calendar tooling. Misuse could lead to dropped long-term tasks.
- Host/process dependency: relies on `nohup sleep` and the underlying machine staying up; no explicit guarantees around reboots, process supervision, or durable job queueing.
- Channel/target correctness: the agent must correctly derive and maintain `--target` values (via TOOLS.md and `session_status`); mistakes can lead to undelivered or misdirected reminders.
- No advanced features: no recurring reminders, editing in place, or rich metadata—changes require removing and re-adding.
**Recommended scenarios**
- Use when the user asks for a one-off reminder within the next 48 hours and does not need calendar-level organization.
- Prefer this skill for quick, high-urgency, low-formality nudges (e.g., oven checks, quick calls, short errands, work follow-ups).
- For anything 2+ days out, or where tracking, rescheduling, or recurring structure is important, route to a calendar skill instead and briefly explain the choice to the user.
Comments (0)
No comments yet. Be the first!