context-budget-audit

Prompt budget auditing tracks per-phase prompt sizes against the model's context window to detect overflow risks. It surfaces sections dominating the prompt over 40 percent and validates whether directives in the prompt were actually covered in the response.

Identify dominant prompt sections by running an audit that flags sections consuming over 40 percent of the prompt. The audit compares per-phase budgets against the model context window and highlights oversized sections suitable for trimming.

Check directive coverage by applying the audit to targeted modules or ranges. The audit validates whether directives in the prompt were actually covered in the response and provides actionable guidance for prompt engineering.

Yes, you can audit specific modules or ranges by providing a track and slug path. The audit applies to the targeted sections and returns metrics including per-phase budgets, oversized sections, directive coverage summaries, and context window headroom.

Context window headroom is the remaining capacity in the model's context window after accounting for prompt sizes. The audit calculates it by comparing per-phase prompt budgets against the model's context window and reports it in the structured metrics.

Build prompts fail to map directives to outputs when directive coverage across phases is incomplete. Auditing the write and wiki sections identifies gaps where directives in the prompt were not actually covered in the response, providing guidance for prompt engineering.