Skip to content

Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.

ci: enhance PR review with severity classification and checklist (#437)

Qin Haojiea1de3ad751f1
diff --git a/.github/workflows/pr-review.yml b/.github/workflows/pr-review.yml
index 06586855d..2b42ab279 100644
--- a/.github/workflows/pr-review.yml
+++ b/.github/workflows/pr-review.yml
@@ -25,3 +25,17 @@ jobs:
github_action_config.auto_review: "true"
github_action_config.auto_describe: "false"
github_action_config.auto_improve: "true"
+ pr_reviewer.extra_instructions: >
+ Classify each finding with a severity tag:
+ [Bug] = Must fix: behavioral regression, logic error, security vulnerability, data loss/corruption, resource leak, API contract violation.
+ [Suggestion] = Recommended: missing tests, dead code, inconsistency, poor observability, unclear intent, unrelated changes, performance issues (N+1 queries, redundant computation in loops, unnecessary large object copies, unbounded collection growth).
+
+ Review focus areas:
+ 1. Control Flow - are all branches reachable? errors propagated correctly? unintended fall-through?
+ 2. Data Flow - input validated at boundaries? implicit type coercion? null/empty handled?
+ 3. Integration Points - API changes backward compatible? config defaults sensible? dependencies pinned?
+ 4. Concurrency - shared state synchronized? execution order assumptions valid? cleanup in all paths?
+
+ Be specific: reference exact variable names, function calls, conditions.
+ When suggesting a fix, include a code block with the recommended change.
+ Match the language of the PR (Chinese PR = Chinese comments, English PR = English comments).