docs: define autonomy-first human-control architecture - #19
Conversation
|
Caution The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased. |
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
Caution The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased. |
What changed
Product architecture
PureFlow now defines three independent outputs beside autonomous code:
Agents may still write every implementation line. The Takeover Envelope exposes where agent capability exceeds current human takeover evidence without blocking normal autonomy or inventing a developer score.
Why
Diff review, generated explanations, quizzes, and plan approvals do not put the developer back into the causal loop. This architecture preserves artifact accountability, behavioral control, and real engineering decision authority while keeping production agents off the human-practice critical path by default.
Validation
git diff --check: pass10339fe: extension, extension-windows, contract, web, jules-rnd-policyHonest gate status
ADR-007, ADR-008, and ADR-009 are Proposed. This PR does not start R5/R6, implement the v0.3 runtime, or claim skill retention. Two independent R7 expert ratings and adjudication remain required before implementation.