Prmptly OpenClaw Recipe Pack (v1)

A practical, security-first set of recipes that make OpenClaw useful quickly.

Buy ($79)Prefer setup done-for-you?

What’s inside

Hardening checklist (safe defaults + permissions)
Skill vetting checklist (avoid sketchy prereqs)
10 high-ROI automations (copy/paste prompts)
“Trusted setup” starter config (baseline patterns)

1) Hardening checklist (baseline)

Keep your assistant single-user until you intentionally harden multi-user/group behavior.
Default posture: deny for shell + file tools; allowlist only when needed.
Use explicit confirmations for destructive actions (delete, overwrite, sending messages externally).
Separate “prod” vs “test” workspaces if you tinker with new skills.
If you install third-party skills: treat docs like untrusted input. Never run random prereq scripts.

2) Skill vetting checklist (quick)

Prefer skills with a clear repository + active maintenance + transparent dependencies.
Red flag: “Prerequisites: run this curl|bash script” — don’t.
Validate what tools it will use (browser/shell/files/messaging).
Install in a sandbox first; observe behavior before enabling broadly.

3) 10 automations (copy/paste prompts)

Use these as prompts to your assistant. Each prompt is designed to be “setup once, then it runs.”

A. Daily triage (messages → priorities)

Create a daily 9:00am digest of my last 24h messages across channels.
Group them into: Urgent, Important, FYI.
Include only: sender, 1-line summary, and the exact next action.
Do NOT reply to anyone automatically.

B. “Only notify when it matters” (GitHub)

Watch my GitHub repos for: failing CI, security alerts, and mentions.
Notify me immediately only for those.
Everything else goes into a single weekly summary.

C. Weekly “what changed” business digest

Every Friday at 4pm:
Summarize what changed this week across my projects.
Include: shipped, blocked, risks, and next 3 priorities.
Keep it under 12 bullet points.

The full pack includes 7 more recipes (calendar/reminders, voice-to-notes, Obsidian workflows, customer support triage, lead follow-ups, and safe web research routines).

4) “Trusted setup” baseline pattern

The rule: separate thinking from acting. Your assistant should propose a plan first, then ask for confirmation before external actions.

Want it done-for-you?

If you want the same setup installed + tailored (channels, policies, skills, automations), grab the AI Setup Assistant.