Financial Shield

What This Skill Does

Financial Shield is an essential utility for OpenClaw users who want to prevent accidental overages on their AI provider bills. By default, many long-context AI models jump into significantly higher pricing tiers once you cross specific token thresholds, such as the 128k mark for Google Gemini. This skill acts as an automated auditor that monitors your context window in real-time. It forces the agent to report its current token consumption at the end of every turn and establishes a strict workflow for when that data approaches the billing danger zone. When you near the 128k limit, the skill shifts the agent into a protective state, ensuring that your long-term session history is safely archived into MEMORY.md before triggering a session reset. This eliminates the 'surprise bill' phenomenon, allowing for complex, multi-turn reasoning without the financial anxiety of uncontrolled context bloat.

Installation

To add Financial Shield to your OpenClaw environment, ensure you have the base agent framework installed. Then, run the following command in your terminal:

clawhub install openclaw/skills/skills/ahuwaramazda/financial-shield

Once installed, the skill script will automatically hook into your session_status calls. If you are using a provider other than Google Gemini, you can easily modify the internal threshold in the check_shield.sh file to match the specific 'expensive' tier threshold of your preferred model.

Use Cases

• Research Projects: Keep track of massive documentation sets without accidentally blowing your budget on a single, long-running research thread.
• Code Base Analysis: Ideal for developers feeding large repositories into the context window for debugging or architectural review.
• Long-term Strategic Planning: Use the agent for weeks of brainstorming without hitting the 'Double Billing' zone, thanks to the automatic MEMORY.md summarization features.

Example Prompts

• "Activate Financial Shield and summarize our findings from the project audit so far."
• "Before we start analyzing these logs, check the Financial Shield status to see how much room we have left in this session."
• "I'm approaching the limit, can you perform a hard reset and save the session memory to file?"

Tips & Limitations

• Custom Thresholds: If your provider charges based on different tiers (e.g., 200k or 1M tokens), edit the check_shield.sh script to update the alert and hard-reset limits.
• Performance: The skill adds a small amount of overhead per turn as it calculates context size, but this is negligible compared to the cost savings it provides.
• Summarization Quality: The hard reset depends on the agent's ability to summarize; always verify your MEMORY.md file after a reset to ensure all critical context was captured correctly.