Skip to content

Turn any idea, plan, or codebase into a beautiful interactive diagram. An agent skill for Claude Code, Codex, and more.

chore(review): reduce repeated CodeRabbit evidence requests (#394)

tt-a1i6db72a9aea3d
diff --git a/.coderabbit.yaml b/.coderabbit.yaml
index f6972c2a..e46a6cc3 100644
--- a/.coderabbit.yaml
+++ b/.coderabbit.yaml
@@ -10,10 +10,14 @@ reviews:
request_changes_workflow: false
high_level_summary: true
high_level_summary_in_walkthrough: true
+ changed_files_summary: false
high_level_summary_instructions: >-
- Summarize the user problem, changed behavior, and compatibility impact.
- Distinguish author-reported validation from observed checks at the current
- head. Do not claim browser or perceptual acceptance from static tests.
+ In at most 100 words, summarize the user problem, changed behavior, and
+ compatibility impact without a file-by-file inventory. State the reviewed
+ base/head and distinguish author-reported validation, reused evidence at its
+ original revision, and observed CI at its checked revision. Treat the summary
+ as a snapshot, not a live CI status. Do not claim browser or perceptual
+ acceptance from static tests.
poem: false
in_progress_fortune: false
sequence_diagrams: false
@@ -47,11 +51,13 @@ reviews:
unrelated rebuilds or infer correctness from generated file volume.
Use the PR base-to-head diff; a clean checkout does not mean an empty PR.
On revision, review new changes and unresolved findings, expanding only
- when intervening changes invalidate earlier evidence. Label feedback as
+ when intervening base, scope, or code changes invalidate the agreed scope
+ or earlier evidence. Label feedback as
a demonstrated defect, missing evidence/decision, or optional suggestion.
Give each request an affected contract, smallest remedy, and responsible
- role (author or maintainer); consolidate requests already answered in the
- PR body, CI, or discussion and retain the evidence's original revision.
+ role (author or maintainer). Before repeating a request, read the latest
+ PR body, CI, and discussion; cite the remaining gap rather than repeating
+ an answered checklist. Retain reused evidence's original revision.
- path: "archify/**"
instructions: >-
Trace callers of changed shared code across diagram types and public
@@ -90,10 +96,13 @@ reviews:
require a PR to close every item of a tracking issue. Accept
equivalent prose and explained Not applicable entries, not just exact
headings or checked boxes. Warn once with the specific missing facts;
- do not classify missing evidence as a confirmed runtime defect.
+ read the latest description and maintainer decisions before repeating
+ a warning. Do not classify missing evidence as a confirmed runtime defect.
- name: Validation evidence
mode: warning
instructions: >-
+ Read the latest PR description, discussion, and CI before evaluating;
+ do not use an earlier bot summary as the source of current evidence.
Apply the change-specific evidence requirements in CONTRIBUTING.md
and the PR template. Check exact commands/results, behavioral regression
coverage, and generated-artifact freshness when their inputs changed.
@@ -107,12 +116,21 @@ reviews:
an excluded binary diff alone is not evidence of unrelated changes.
Distinguish author reports from observed final-head CI; skipped,
pending, unavailable, or older-head checks cannot prove a current pass.
+ For revisions, identify which changed behavior invalidates earlier
+ local evidence before requesting reruns. Reuse unaffected evidence
+ with its original revision and reuse rationale; a new head alone does
+ not require repeating every local check. Required final-head remote
+ checks still apply. If current evidence cannot be read, report that
+ access limitation rather than claiming the author did not supply it.
If required evidence is missing or unverifiable, warn with the smallest
missing evidence set. Do not require screenshots for non-visual changes
or infer a browser pass from a unit test. Accept explicit maintainer
exceptions and scope-appropriate explanations for omitted checks.
When fork CI awaits approval, identify maintainer action after inspecting
the workflow changes; do not ask the author to obtain unavailable rights.
+ State the evaluated head and evidence revisions in the result. Consolidate
+ remaining gaps with their affected contract, smallest check, and owner;
+ do not repeat satisfied requirements or unrelated rerun instructions.
A bot pass or exception never waives required CI or branch protection.
knowledge_base:
code_guidelines:
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index c7df5482..e3ed3162 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -130,6 +130,13 @@ instead of pushing an empty commit to retrigger the bot:
| Several rapid revisions are in progress | `@coderabbitai pause`, then `@coderabbitai resume` when ready |
Check the updated summary for results; a command acknowledgment is not completion.
+Treat its evidence assessment as a snapshot at the stated revision. After updating
+only the description or after CI completes, use the pre-merge command above to
+refresh stale results; editing the description is not proof that checks reran.
+Before repeating a bot request, read the latest evidence and identify the remaining
+gap. Reuse unaffected local results with their original revision and a reuse
+rationale; rerun checks whose evidence the new changes invalidate. Required remote
+checks must still pass on the final head.
If fork CI needs approval, a maintainer must inspect the proposed workflow/code
and handle the GitHub approval. Authors should link the waiting run and continue
checks available to them; they are not expected to grant themselves CI access.