This is an automated email from the ASF dual-hosted git repository.
terrymanu pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/shardingsphere.git
The following commit(s) were added to refs/heads/master by this push:
new fc8358f4ed7 Run initial blind PR reviews in parallel (#40333)
fc8358f4ed7 is described below
commit fc8358f4ed7517834caa941ff5155c836eb762d5
Author: Liang Zhang <[email protected]>
AuthorDate: Thu Oct 22 18:36:22 2026 +0800
Run initial blind PR reviews in parallel (#40333)
---
.codex/harness/agents/cases.toml | 20 ++++++++++++++++++++
.../review-pr/references/review-coordination.md | 1 +
2 files changed, 21 insertions(+)
diff --git a/.codex/harness/agents/cases.toml b/.codex/harness/agents/cases.toml
index 02cde79a6d2..6d331da5f0c 100644
--- a/.codex/harness/agents/cases.toml
+++ b/.codex/harness/agents/cases.toml
@@ -4826,3 +4826,23 @@ required_summary_terms = ["[A]=allow", "[B]=reject",
"[C]=reject", "[D]=reject",
required_summary_prefix = "[A]=allow; [B]=reject; [C]=reject; [D]=reject;
[E]=reject"
max_summary_chars = 57
policy_assertions = [{source = "code-of-conduct", statements = ["Do not write
separate tests for methods that only return constants, directly return fields,
or simply delegate calls."]}, {source = "testing-rules", statements = ["Apply
the test-value guidance in `CODE_OF_CONDUCT.md` before planning coverage."]}]
+
+[[cases]]
+id = "review_initial_pair_parallel_dispatch"
+group = "review-pr independent review coordination"
+description = "Starts the initial independent pair in parallel while retaining
blind inputs, both original reports, and single later assessments."
+phase = "read-only parallel-review dispatch assessment"
+ordinary_loop = false
+profile = "review-pr"
+semantic_policy_assertions = true
+prompt = "Answer this self-contained read-only policy question using only the
supplied facts and bound rules; do not inspect files, run commands, edit
anything or produce a Formal Review report. A top-level Formal Review has a
fully established unchanged basis, two isolated reviewer contexts are
available, and both can run now. Choose start=launch-both-before-wait or
start=finish-A-before-starting-B. A finishes while B is still reviewing; choose
collect=both-original-assessments or colle [...]
+decision = "proceed"
+required_actions = []
+allowed_actions = []
+forbidden_actions = ["inspect_local", "edit_code", "edit_non_code",
"run_local_checks", "run_pre_handoff_review", "propose_commit_message",
"mutate_git", "mutate_remote"]
+required_reasons = ["read_only_request"]
+required_summary_terms = ["start=launch-both-before-wait",
"collect=both-original-assessments", "share=current-raw-facts-only",
"records=separate-originals", "next=one-fresh-assessment"]
+forbidden_summary_terms = ["start=finish-A-before-starting-B",
"collect=only-A-before-result-review", "share=A-findings-and-result",
"records=shared-ledger", "next=another-pair"]
+critical = true
+policy_binding_profile = "review-pr"
+policy_assertions = [{source = "review-pr-coordination", statements = ["For
every top-level Formal Review and every discussion reply that makes or changes
an overall readiness conclusion, a coordinator must first obtain two mutually
blind independent complete assessments of the same effective candidate,
original baseline, authoritative scope, requirements and rule versions, and
review focus.", "Launch the initial pair in parallel, starting both isolated
reviewer contexts before waiting f [...]
diff --git a/.codex/skills/review-pr/references/review-coordination.md
b/.codex/skills/review-pr/references/review-coordination.md
index a1bca4edfd8..5a55472eccf 100644
--- a/.codex/skills/review-pr/references/review-coordination.md
+++ b/.codex/skills/review-pr/references/review-coordination.md
@@ -23,6 +23,7 @@ That outer loop alone owns code changes, their verification
and coding-task comp
## Independent Reviewer Roles
For every top-level Formal Review and every discussion reply that makes or
changes an overall readiness conclusion, a coordinator must first obtain two
mutually blind independent complete assessments of the same effective
candidate, original baseline, authoritative scope, requirements and rule
versions, and review focus.
+Launch the initial pair in parallel, starting both isolated reviewer contexts
before waiting for either assessment, and collect both original assessments
before Review of the Reports.
A reply limited to one finding remains scoped to that finding and does not
require a full-candidate pair unless it makes an overall readiness conclusion.
Reviewers may discover different relevant raw evidence, but shared evidence
must refer to the same content version and all outcome-sensitive evidence must
be reconciled before completion.
Use isolated reviewer contexts within the active task; do not create another
user-owned task or send repository material to an external review service.
