deep-research-orchestrator

What This Skill Does

The deep-research-orchestrator is a specialized agent skill designed to bridge the gap between high-level business goals and granular advertising execution. It acts as an orchestrator that deconstructs complex growth questions into testable hypotheses, gathers evidence from various advertising platforms (Meta, Google, TikTok, Amazon, and DSPs), and synthesizes this data into actionable, channel-specific strategies. Unlike general-purpose research tools, this skill is purpose-built to handle the unique constraints of ad ecosystems, ensuring that recommendations are grounded in metrics like ROAS, CPA, and budget efficiency while respecting the specific operational nuances of each platform.

Installation

To integrate this skill into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/danyangliu-sandwichlab/deep-research-orchestrator

Use Cases

This skill is ideal for performance marketers, growth leads, and ad operations teams facing critical decision points. Use it to:

• Analyze why ROAS is dipping in a cross-channel campaign and receive a step-by-step mitigation plan.
• Evaluate if budget reallocation between Meta and Google Ads will result in lower acquisition costs.
• Orchestrate complex A/B testing frameworks across Amazon and TikTok Ads based on current creative performance.
• Determine the optimal bid strategy for a new product launch based on industry benchmarks and current account history.

Example Prompts

1. "I need to decide if we should shift 20% of our Meta budget to TikTok Ads by next Friday. Create a research plan to evaluate the impact on our overall ROAS."
2. "My Google Ads CPA has increased by 15% over the last week. Orchestrate a research workflow to identify if this is a seasonal trend or a decline in audience quality."
3. "We are launching a new video campaign on YouTube and Meta. Develop a hypothesis set for creative performance and a plan to monitor and iterate based on real-time metrics."

Tips & Limitations

To get the best results, ensure your 'research_question' is specific and time-bound. Always provide a clear 'decision_deadline' to allow the orchestrator to adjust its evidence-gathering depth. Note that this skill is an advisory tool; while it provides high-signal strategies, it does not execute changes directly. Always review the 'Actionable Next Experiments' output before pushing changes to your live ad accounts, and ensure you have sufficient data integrity in your tracking systems before running an analysis.