docs: give a hand-driven PR a stopping point, and name the stacked-PR blind spot

CONTRIBUTING.md

@@ -312,7 +312,7 @@ This repo uses AI reviewers and an autonomous fix loop — full details in the d
 [AI-Assisted Development](https://bestax.io/docs/guides/getting-started/ai-development).
 The short version for contributors:
 
-- **Every PR gets a CodeRabbit review** automatically. Address or refute its comments — it
+- **A PR targeting `main` gets a CodeRabbit review** automatically once it is out of draft. Address or refute its comments — it
   reviews incrementally and marks addressed comments "✅ Addressed". It also rate-limits, so a
   push during a spent window waits for the next one; the AI-assisted section below says how to
   nudge it. A human maintainer still reviews and merges everything.
@@ -330,6 +330,18 @@ The short version for contributors:
   its window reopens.
   Do not relabel per push: each application spends a full opus session, and relabeling after
   every fix is what turned #643 into 14 review rounds.
+- **Know when to stop fixing.** The autonomous loop stops after `MAX_ITERATIONS` fix rounds
+  (`claude-pr-loop.yml` sets it), pauses, and labels the PR `ai-loop-paused` for a human. Give
+  a hand-driven PR the same stopping point. Once a file has been round-tripped that many times
+  and what is left is wording rather than behaviour, reply once naming the findings you are
+  leaving and why, then take the PR to human review. The reviewers keep reviewing; what changes
+  is that you stop treating every finding as blocking. A finding that names a defect in
+  behaviour still earns another round.
+- **A PR based on another PR's branch gets no automatic CodeRabbit review.** It auto-reviews
+  only the base branches listed in `.coderabbit.yaml`; ask for one with `@coderabbitai review`.
+  The deep review turns on its own condition — a branch whose copy of `claude-review.yml`
+  differs from the default branch's does not get one — so an upper PR carrying a workflow change
+  can lose both. Prefer landing a stack one PR at a time; the docs guide has the detail.
 - **PR titles must be scoped conventional commits** — the title becomes the squash commit and
   drives semantic-release (see [Commit Message Guidelines](#commit-message-guidelines)).
 

docs/docs/guides/getting-started/ai-development.md

@@ -30,15 +30,27 @@ The more actionable your issue, the better the agent does with it:
 
 ## For contributors: what reviews your PR
 
-Every PR gets AI review before human review:
+AI review comes before human review, and which reviewers a PR draws depends on where it
+targets and what it changes:
 
-- **CodeRabbit** reviews automatically. Respond in-thread or just push fixes — it re-reviews
+- **CodeRabbit** reviews a PR automatically once it targets the default branch and is out of
+  draft. Respond in-thread or just push fixes — it re-reviews
   each push and marks addressed comments with "✅ Addressed in commit …". If you think a
   finding is wrong, say so in the thread; a maintainer has the final word.
 - **`@claude` mentions** are restricted to the owner, members, and invited collaborators (they
   spend the maintainer's Claude usage). External contributors don't need them — just push.
 - **Copilot review** may also appear when the maintainers have it enabled.
 
+**A PR based on another PR's branch gets no automatic CodeRabbit review.** CodeRabbit
+auto-reviews only the base branches listed under `auto_review.base_branches` in
+`.coderabbit.yaml`, which is the default branch alone. On any other base it posts a "Review
+skipped" notice, and that notice names `@coderabbitai review` as the way to get a single review
+anyway. Stacking affects that reviewer only. The Claude deep review turns on a separate
+condition — a branch whose copy of `claude-review.yml` differs from the default branch's does not
+get one — which an upper PR carrying an unmerged change to that file meets, so the two are easy
+to confuse. Mention CodeRabbit on an upper PR, and prefer landing a stack one PR at a time so
+each is reviewed against the default branch.
+
 A green AI review is not approval: a human maintainer still reviews and merges every PR, and
 the review-time requirements (Storybook story for UI changes, docs page for API changes,
 `skills/` updates for component changes) still apply.