Buddy Followup

Agents say "I'll follow up" — and then forget. This skill fixes that. When you kick off a long-running task (sub-agent, build, API call, script), set a timer...

Security Scan

The skill's code and instructions are consistent with its stated purpose: it schedules follow-ups by calling the OpenClaw CLI to create per-channel cron jobs and does not request unrelated credentials or perform unexplained actions.

Purpose & Capability

Name/description match the implementation: the script parses a delay and task text, reads channel configuration via the openclaw CLI, and uses openclaw cron add to schedule per-channel follow-ups. Nothing required by the skill (no extra env vars, no external downloads) appears unrelated to implementing follow-ups.

Instruction Scope

SKILL.md instructs the agent to run the included shell script. The script only queries OpenClaw config and calls openclaw cron add to schedule messages; it does not read arbitrary files, exfiltrate data, or call third-party endpoints outside OpenClaw. It does rely on the openclaw CLI and will cause scheduled agent messages to be sent using your configured channels.

Install Mechanism

No install spec is provided (instruction-only plus a small script). No downloads or third-party packages are fetched by the skill itself. The optional npx clawhub install suggestion is external to the skill bundle and not an automatic install step here.

Credentials

The skill requests no environment variables or credentials. It uses the OpenClaw CLI, which in turn will use whatever channel credentials are already configured in your OpenClaw instance — this is proportionate to a skill that needs to post follow-up messages to those channels.

Persistence & Privilege

The script creates temporary cron jobs via openclaw cron add (with --delete-after-run). While not a platform-level privilege escalation, it does modify OpenClaw's scheduled-job state and will cause the agent 'main' to run at the scheduled times. This is expected for a follow-up scheduler but worth noting because it results in future autonomous actions (posting messages) using your configured channels.

Assessment

This skill appears to do exactly what it says: schedule follow-up timers by using your OpenClaw configuration and cron system. Before installing or running it, check the following: (1) ensure the openclaw CLI is installed and you trust the configured channels and the agent named 'main' because scheduled jobs will post on those channels; (2) avoid putting sensitive secrets or private data into the TASK string since that text will be sent to your channels when the timer fires; (3) inspect the included scripts (scripts/followup.sh) yourself — it's short and readable — and test on a non-production channel first to confirm behavior; (4) note operational caveats (date -d use and python3 for JSON parsing may behave differently on some systems). If you want extra safety, run with a test channel or modify the script to require explicit confirmation before scheduling.

Agents say "I'll follow up" — and then forget. This skill fixes that.

When you kick off a long-running task and tell the user "I'll get back to you", use this to actually follow through — automatically, on every configured channel.

Installation

Via ClawHub (recommended):

npx clawhub install buddy-followup

When to Use

• You launch a sub-agent, script, build, download, or API call that takes time
• You tell the user "I'll update you when it's done" or "give me a few minutes"
• You need to check back on something after a delay

How It Works

1. You estimate how long the task will take and call the script with that delay
2. Cron jobs fire at the given time on every configured channel (Telegram, WhatsApp, etc.)
3. You wake up with the task context, check status, and reply to the user
4. Still pending? Set another timer and keep the loop going
5. Done? Report results and close the loop

No hardcoded IDs — channels and targets are read dynamically from openclaw config get at runtime.

Usage

bash ~/clawd/skills/buddy-followup/scripts/followup.sh <delay> "task context"

Delay formats: 30s, 5m, 2h

Examples:

bash ~/clawd/skills/buddy-followup/scripts/followup.sh 2m "check if sub-agent finished building the API"
bash ~/clawd/skills/buddy-followup/scripts/followup.sh 10m "check if deployment completed"
bash ~/clawd/skills/buddy-followup/scripts/followup.sh 30s "verify test results are ready"

Run via exec tool — the script exits immediately after scheduling. No background flag needed.

When the Follow-Up Fires

On receiving ⏰ FOLLOW-UP (<delay>): <task>:

1. Check the task status
2. Reply directly — routes to all configured channels automatically
3. Done → confirm results to the user
4. Still running → tell the user, reset the timer:

   bash ~/clawd/skills/buddy-followup/scripts/followup.sh 2m "still waiting for X, checking again"
   

Requirements

• OpenClaw gateway running
• At least one channel configured (Telegram with channels.telegram.defaultTo, or WhatsApp with channels.whatsapp.allowFrom)
• openclaw CLI available in PATH

Notes

• Channels are discovered at runtime — adding a new channel automatically includes it
• Each timer creates one cron job per channel, deleted after firing
• The agent decides the delay — base it on realistic task completion time