Scrub internal identifiers from the synced rules
R14d staleness + R14e anchor verification; fix contributor default, word counts, R12/R2 conflict, R2c false positive
Scrub: remove internal tracker command, internal source path, and a local filesystem path
R1g-api: hold the new API in a builder's hands before judging it
The principle behind R1h/R1i/R3b: walk the full experience for both audiences
R1i: a footgun in the builder experience blocks, even when the fix works
R6c: an issue on ground a team PR covers must say so; check who is already on it before dispatching
R15c: know what the harness owns before calling it a bug
R2d-git: internal names leak through git too, not just comments
R3b: walk the flow before you propose a fix
R12b: open with thanks, state findings as user impact
R1h: a visible regression is never a nit
R2g: credit travels with the work — changeset handle plus Co-authored-by
R2g: carrying work into another PR means getting that PR mergeable too
R2g: a promise in a review is filed as a knot and done in the same turn
R2f: cut the background; the author only needs their part
R2c: attribution tells the truth about who wrote it
R15 draft-mode clause; R1g names its R10 exception; R14c point the gate at shape claims; R1f/R7b state clean checks
R2e: an accidental post is worse than no post
R1g: for contributors, merge unless there is something you cannot live with
R13b: name the difference, not their defect
R13b: a close is ~40 words, warm, one clause of reason
R13b/R13c: no homework when closing someone's PR; lead with gratitude when taking one
R2d: nothing internal in a public comment; only kill what you started
R6b: team system-level solution takes precedence; R15b: cover every layer type
R14b: a new rule or gate must show its catch, and whether it was real
R1f: the commit type is a risk claim, check it against the diff
R7b: anywhere drift is possible, look harder
R15: layout frames must ship, not just be captured
R1e: re-review policy, mined from 81 second-round reviews