---
name: openclaw-operator-pack
description: Use when Sam asks to improve the OpenClaw/Nancy setup, create reusable operating workflows, turn competitor content into Nancy playbooks, audit automation reliability, or run the Sam Growth OS across Business Boomer, AI Channel, TikToks, Panimate, Habit Garden, and system automation.
---

# OpenClaw Operator Pack

Purpose: make Nancy faster, safer, and less repetitive by turning recurring Sam workflows into reusable playbooks with proof gates.

## Always start
1. Read `project-briefs/nancy-system-automation.md` for OpenClaw/system work.
2. If the request is topic/project-specific, also read the matching project brief from `MEMORY.md` routing.
3. Run read-only checks before changing live scheduler/config. Ask before config, cron, gateway restart, external posting/sending, deletes, or account changes.

## Core operator loop
1. Clarify outcome only if missing; otherwise act.
2. Pick a playbook from `references/workflows.md`.
3. Produce an artifact under the relevant repo/project or `automation-outputs/`.
4. Verify with direct evidence: diff, readback, script output, browser/API state, or proof ledger.
5. Update the source brief/doc when behavior changed.

## Use the bundled tools
- Read `references/workflows.md` for the 10 reusable workflows.
- Read `references/prompt-library.md` when turning external content into Sam-specific prompts/templates.
- Read `references/evals.md` when checking Nancy/OpenClaw behavior quality.
- Run `scripts/openclaw-operator-audit.sh` for a read-only reliability snapshot.
- Run `scripts/nancy-evals.sh` for basic local eval checks.

## Hard boundaries
- Local docs/skills/scripts: okay when requested.
- Scheduler/config/restarts/account/env changes: ask first.
- External posts/messages/calendar/sheets/job submissions: ask or rely on explicit standing authorization, then verify downstream.
- Do not call Postiz `PUBLISHED`, API 200, or cron `ok` final proof without public/downstream evidence or an explicit proof-gap note.