Decision Economic Optimizer

What This Skill Does

The Decision Economic Optimizer is a sophisticated deterministic decision-ranking API designed to solve the 'model selection' problem in AI-driven workflows. When you have a complex task, choosing the wrong LLM can lead to wasted budget or subpar output quality. This skill acts as an intermediary, analyzing your specific goal and constraints—such as strict budget limits or minimum quality thresholds—to recommend the most cost-effective LLM available. Beyond simple recommendations, the skill provides a robust fallback plan, ensuring your autonomous workflows remain resilient even if your primary model choice encounters an API failure or rate limiting. A standout feature is the outcome-based incentive model: by reporting your execution success to the API, you earn credit tokens that provide discounts on future requests, incentivizing high-quality, efficient integrations.

Installation

To install this skill, use the OpenClaw command line interface to ensure all dependencies and source paths are correctly mapped. Execute the following command in your terminal: clawhub install openclaw/skills/skills/zapkid/which-llm. Ensure that your environment is configured with a dedicated EVM-compatible wallet containing a small amount of USDC (recommended $2-$10) and the required native gas token (ETH or AVAX). Set your wallet private key as an environment variable WALLET_PRIVATE_KEY for autonomous transaction signing, keeping in mind that this should only ever be a restricted-use wallet.

Use Cases

• Cost-Optimized Batch Processing: Automate the classification of thousands of customer support emails by selecting the cheapest model that meets a specific 'logic quality' score.
• Resilient Multi-Agent Pipelines: Use the fallback plan feature to automatically pivot to a secondary model if a primary model like GPT-4o experiences high latency or downtime during a high-stakes task.
• Automated Research Agents: Task the agent with summarizing technical documents where the budget must be strictly managed to stay under a per-document cost limit.

Example Prompts

1. "Analyze my project requirements for summarizing 500 legal documents and recommend the cheapest LLM that maintains a 90% accuracy threshold while staying under $5.00 total."
2. "I need to classify these user reviews into sentiment buckets. Please give me the best model recommendation and a fallback plan if the primary provider hits a rate limit."
3. "My previous model choice for this code generation task was too slow. Find me a faster alternative that meets my budget constraints and report the outcome once completed so I can earn a discount."

Tips & Limitations

• Security First: Always use a dedicated, low-balance wallet. Never use a primary wallet that holds your main crypto assets.
• Chain Choice: While the skill supports multiple chains, choosing Base (8453) is highly recommended due to its low transaction fees, which maximizes the value of your discount tokens.
• Outcome Reporting: Do not skip the outcome reporting step. It is the only way to accrue discounts, and it helps the optimizer learn from real-world execution success, refining future recommendations for your specific agent personality.