auditing-commitmentslisted

# Auditing Commitments A proposal is a binding offer. Every "we will", "we shall", "we guarantee" and "we ensure" is a candidate obligation, and if it survives into the contract the offeror performs it at its own cost for the life of the award — sometimes with the proposal incorporated wholesale by reference. Overcommitting is not a writing problem, it is a margin problem that surfaces in month three of performance. This skill inventories every promise in a draft, maps each one to the requirement that justifies it, and prices the difference where the promise is stronger than the requirement. ## When to use this skill Use this skill after drafting and before the Red Team, on any volume that makes performance commitments. Run it again on any section rewritten during review recovery. Run it always on **AI-generated or AI-assisted draft content**. Generated proposal prose overcommits substantially more than human prose, and it does so in one direction: fluent future-tense verbs read as confidence, so a model optimising for confident prose accumulates obligations. The draft sounds strong and the delta is invisible until award. Distinguish it from adjacent skills: - `checking-submission-compliance` asks whether the proposal will be *accepted*. This skill asks what the proposal *obligates you to do*. - `reviewing-color-teams` asks whether the proposal is persuasive and scoreable. This skill is indifferent to persuasion; a beautifully written unbounded guarantee is exact