Cost governance

Prompt optimizer

Spot duplicate or conflicting instructions, compress context, and split heavy work into phases — before you spend.

Length

≈ 1,820 tok

Most builder tasks: under 1,200

Estimated savings

≈ 35%

If suggestions are applied

Risk flags

2

Conflicting + duplicated guidance

Prompt length

Info

Your prompt is ~1,820 tokens. Most builder tasks complete well under 1,200.

≈ 35% tokens

Repeated instructions

Warning

You ask to 'keep design tokens only' twice. Merging will reduce noise.

≈ 90 tokens

Conflicting guidance

Warning

You request both a minimal change and a broad refactor. Pick one scope.

Split into phases

Tip

This work covers UI, data and routing. Splitting reduces risk and cost.

≈ 22% spend

Compress shared context

Tip

Reuse a project memory entry instead of re-pasting style rules every time.

≈ 180 tokens / message

Optimized prompt preview

The optimizer will produce a tighter version preserving your intent. Foundation only — the live engine ships in a future phase.

[Preview placeholder] Concise instructions, merged style rules, scope limited to the requested change. Estimated tokens: ≈ 1,180.

Why this matters

  • • Shorter prompts cost less and finish faster.
  • • Conflicting instructions cause re-runs — the most expensive failure mode.
  • • Phase-aware prompts let the scheduler fit work into your weekly budget.

Pairs well with

  • Budget mode: Balanced
  • Routing: GPT-5 for code
  • Scheduler: Split into phases