# Working Locally With No-Push Intent Limits AI Contribution Risk ## Claim Running an AI coding tool locally with an explicit no-push intent avoids the review-burden and supply-chain risk of AI-authored pull requests. <!-- atomicity: 1 idea ✓ --> ## Scope & conditions - Holds for exploratory/prototyping use where output is discarded or hand-carried by a human into their own commit. - Doesn't apply to any workflow where the AI's output is committed or pushed directly — that reintroduces the review and provenance risk this posture avoids. ## Evidence Anne McCarthy's workflow: Claude in terminal plus CraftAgents, working a PR locally against a pre-set `wp-env` dev environment, with explicit intent never to push or open a competing PR. The prototype exists only to be shown, then discarded. ## Links - [[AI & the Future of Knowledge Work MOC]] — a bounded way to use agents that doesn't spend anyone else's review capacity - [[3. Resources/Permanent Notes/AI Prototypes Communicate Intent Better Than Competing Pull Requests]] — because applies-in; the no-push posture is what makes the prototype-to-communicate pattern safe. - [[3. Resources/MOCs/Open Source Governance MOC]] — because applies-in ## Tests / falsifiers - If contributors start pushing "prototype" branches directly, this safety posture no longer holds — worth watching whether the norm drifts as the pattern spreads. - Compare review-burden metrics (time-to-merge, review rounds) on threads that used this local-prototype pattern vs. threads with direct AI-authored PRs. ## Maintenance - Last tended: 2026-08-16 - Next review: 2026-11-16 ## Source - [[3. Resources/Literature Notes/Back and Forth — AI as a Communication Device in WordPress Contribution]] — describes the exact local, no-push workflow.