AI Requirement Lock

The problem

AI agents often suffer from "requirement drift," where later instructions cause them to forget earlier constraints or overwrite specific values you needed to preserve. This leads to broken code, lost data, and cycles of frustrating manual corrections.

What it does

• Extracts explicit constraints into a persistent contract of Must Do, Must Not Do, and Preserve Exactly requirements.
• Maintains requirement state across multi-turn conversations to prevent context window decay from dropping old rules.
• Detects and flags logical conflicts between new prompts and existing locked constraints instead of guessing.
• Performs a final internal audit (Pass/Fail/Unverified) against the original contract before delivering the output.

Why this beats prompting it yourself

Standard prompting relies on the LLM's fluctuating attention mechanism, which naturally degrades as conversations get longer. This skill implements a formal verification logic that treats your instructions as a testable specification, ensuring that creative "improvements" never override your hard constraints.

Use cases

• Refactoring legacy code while ensuring specific API signatures or variable names remain untouched.
• Updating marketing copy while strictly adhering to a "Forbidden Words" list and brand guidelines.
• Modifying technical documentation while locking down specific dates, version numbers, and legal disclaimers.
• Executing complex tasks with multi-step acceptance criteria that must all be satisfied simultaneously.