Compare static system-prompt tokens against your budget to see the percentage used, remaining tokens, and a bloat verdict.
Tokens your fixed system instructions and boilerplate consume. Unit: tokens.
The total token budget you reserve for the system prompt. Unit: tokens.
Budget Used
25%
Budget Used — Percentage %. Above 0 increases with input; compare to benchmarks or goals.
Remaining Tokens
1,500 tokens
Remaining Tokens — Calculated outcome tokens. Derived from your inputs via the displayed formula; use to plan.
Verdict
Healthy — room for task instructions
Verdict — Calculated outcome. Derived from your inputs via the displayed formula; use to plan.
What this means
Divide the static token load by the total budget to get the fraction consumed. The result is shown as a percentage with the remaining tokens and a plain-language verdict. A prompt that spends most of its budget on fixed text leaves little room for the user input and the answer, which usually pushes you into a larger...
System prompts tend to creep: roles, style rules, guardrails, few-shot examples, and formatting instructions pile up until the fixed part of every request costs more than the actual task. This tool measures the ratio of static instructions to your total prompt budget so you can decide what to keep, condense, or move to a smaller specialist prompt.
Formula
used = staticTokens ÷ budget
Worked examples
FAQ
What counts as a static token?
Everything fixed in the system prompt: role definitions, tone rules, output schemas, guardrails, and hard-coded examples. Text that changes per request is not static.
Is a smaller system prompt always better?
Not always. Some instructions improve accuracy even when they cost tokens. Use the check to find the point of diminishing returns rather than trimming to zero.