flow-skill-analyze-context

To estimate total token cost for AI prompts, you can aggregate conversation history, system prompts, active rules, and tool definitions to produce a single token-based estimate for budgeting.

Budgeting token usage across long AI sessions involves aggregating inputs from history, prompts, rules, and tool definitions, then applying a character-based token approximation to calculate the total context cost.

Yes, the token estimation includes system prompts, active rules from AGENTS.md, conversation history, attached context blocks, and tool definitions to account for all factors driving context size.

Character-based token approximation works by counting characters in your aggregated inputs and applying a 0.3 multiplier to yield token counts, providing an adaptive estimation for prompt optimization.

Yes, you can optimize prompts to stay within a specific token budget by running an analysis to estimate how many tokens will be consumed and adjusting your conversation history, rules, and prompts accordingly.

The limitation of using character counting for token estimation is its adaptive nature; it approximates tokens by multiplying characters by 0.3 rather than using an exact tokenizer, which may yield slight variations in cost estimates.