Mechanism: Negative Space — Say What Not to Touch
Negative space is the practice of naming what must stay unchanged — files, behaviors, APIs, and styles — so the…
Mechanism: Rollback — Undo Before the Agent Goes Too Far
Rollback is having a known-good restore point — commit, stash, branch tip, or snapshot — before agent edits land, so…
Mechanism: Checkpoint — Review Before You Merge
A checkpoint is a mandatory pause — human or automated — between agent output and merge, where you review the…
Mechanism: Context Budget — Spend Tokens on What Matters
A context budget is a deliberate limit on what you load into the model's window — files, history, rules, and…
Mechanism: Scope Fence — One Job Per Session
A scope fence is a deliberate boundary around one session so it handles a single job — one goal, one…
Mechanism: Verify Loop — Check Every Pass Before You Trust “Done”
A verify loop is running an explicit check after each agent pass and feeding only the results back, so done…
Mechanism: Termination Condition — How to Stop an AI Loop on Purpose
A termination condition is an explicit, checkable rule that tells a loop or agent run to stop by design —…
Mechanism: Loop — Repeat Until the Job Is Actually Done
A loop is running the same goal-driven pass again — prompt, edit, check — until a defined stop condition holds.
Mechanism: Constraints — Stop the Agent From “Improving” the Wrong Things
Constraints are explicit limits on what the agent may change, add, or assume — stated alongside the goal so helpfulness…
Mechanism: Goal — Define What “Done” Means Before You Prompt
A goal is an explicit statement of what must be true when work is finished — symptom, boundary, and how…