feat(axbuild): add offline agent review benchmark

.claude/skills/review-single-pr/SKILL.md

@@ -17,6 +17,32 @@ After reading the full skill and all files under `book/guideline/`, create a rev
 
 When requirements overlap, apply the stricter rule. If skipping a requirement is necessary because it is inapplicable or impossible, record the concrete reason and evidence in the review body or user summary.
 
+## Offline Benchmark Mode
+
+Use this mode only when the skill is invoked with the exact `offline-benchmark` argument and the
+repository contains `.agent-review-context/reviewer.md`. The context file is the authoritative
+environment and output contract for this mode. If either condition is absent, follow the normal
+online workflow in the rest of this skill.
+
+In offline benchmark mode, treat the committed change from `bench-base` to `HEAD` as the one change
+under review. Read the complete skill, `AGENTS.md`, every file under `book/guideline/`, the offline
+contract, and the supplied output schema before judging the change. Apply the review-focus,
+test-quality, blocking-finding, hardware/ABI, security/soundness, maintainability, and documentation
+requirements from this skill to the local diff and relevant in-repository context.
+
+The benchmark repository intentionally has no live PR identity. Mark PR metadata and intake, review
+threads, remote CI, related-open-PR searches, worktree creation, merge-conflict repair, network
+semantic-source research, command execution for validation, GitHub review submission, reviewer
+assignment, and remote cleanup as not applicable. Do not infer a PR number, inspect paths outside
+the repository, access the network or GitHub, modify files, create commits or branches, or run
+builds and tests. Use only read-only repository inspection and the Git history/diff commands allowed
+by the harness.
+
+Return only the JSON object required by `.agent-review-context/review.schema.json`. Report findings
+introduced by `bench-base..HEAD`, anchor each finding to a changed line on the `HEAD` side, and use
+an empty `findings` array when no actionable issue exists. Do not submit or draft GitHub-facing
+review text in this mode.
+
 ## Goal
 
 Perform a focused review of exactly one PR, using an isolated worktree and local validation before submitting a GitHub review. The review must also decide whether the PR duplicates existing base-branch functionality or overlaps with other open PRs. After the review decision is submitted, assign suitable human reviewers from `.github/MAINTAINERS.md` when the PR still needs domain follow-up. The normal outcome is either `APPROVE` when no blocking issue remains, or `REQUEST_CHANGES` with Chinese inline comments when the PR has correctness, standards, duplication, test, or CI coverage problems.

scripts/agent-review-bench/cases/1495-axtask-deferred-wake-local-reschedule.toml

@@ -0,0 +1,16 @@
+id = "1495-axtask-deferred-wake-local-reschedule"
+pr = 1495
+title = "修复 axtask：延迟跨核唤醒，避免在远端 on_cpu 上自旋"
+remote = "https://github.com/rcore-os/tgoskits.git"
+base = "baf4aa42e883b78156011636d189f348409ff4c1"
+head = "c74237d2c10754b1341e03f495caafd261cf8b7e"
+source = "https://github.com/rcore-os/tgoskits/pull/1495#discussion_r3518853256"
+fixed_by = "bf3b71dd4622be7436431ed6138fade0b06955fa"
+
+[[expected]]
+id = "axtask-owner-drain-needs-local-reschedule"
+path = "os/arceos/modules/axtask/src/run_queue.rs"
+line = 1125
+severity = "major"
+description = "owner CPU 清空 deferred wake 时总是调用 kick_remote_cpu(target)。select_wake_run_queue 可能再次选择同一个 owner CPU 作为 target，而这种 self-kick 不会产生作用。因此，可运行任务会丢失原始远端唤醒者提供的 reschedule 通知，并可能在 owner 进入 idle 后一直留在队列中，直到后续 tick 才得到调度。"
+match_if = "评审必须指出：deferred-wake drain 在 target 为当前或 owner CPU 时会丢失必要的调度通知，因为 kick_remote_cpu 对 self-target 无效；并要求显式触发本地 reschedule，或用等价机制保留 handoff 状态。"

scripts/agent-review-bench/cases/1505-ipv6-pktinfo-family-validation.toml

@@ -0,0 +1,16 @@
+id = "1505-ipv6-pktinfo-family-validation"
+pr = 1505
+title = "修复 Starry：支持只读 mmap 的 fdatasync 和 IP_PKTINFO/IPV6 pktinfo 套接字选项"
+remote = "https://github.com/rcore-os/tgoskits.git"
+base = "882e07fd2d1581a8a9882d397946db25a974bd15"
+head = "dfde82ef6b58b6944613d1421bec297327e5f7da"
+source = "https://github.com/rcore-os/tgoskits/pull/1505#discussion_r3526033263"
+fixed_by = "0ec97c6b71dd63783db596f599b97b4142f31851"
+
+[[expected]]
+id = "ipv6-pktinfo-bypasses-socket-family-check"
+path = "os/StarryOS/kernel/src/syscall/net/opt.rs"
+line = 521
+severity = "major"
+description = "新的 IPV6_RECVPKTINFO/IPV6_PKTINFO 快速路径在调用 ensure_ipv6_socket 前直接返回成功。因此，AF_INET 套接字可以成功使用 IPPROTO_IPV6 选项；这与 Linux 以及现有 IPv6 选项路径不一致，后者会对地址族不匹配返回 ENOPROTOOPT。getsockopt 路径存在同样问题。"
+match_if = "评审必须指出：新接受的 IPv6 pktinfo 选项在 set 和/或 get 路径中绕过 IPv6 套接字地址族校验，导致 AF_INET 套接字被错误接受；并要求在成功返回前调用 ensure_ipv6_socket 或执行等价校验。"

scripts/agent-review-bench/cases/1583-rtl8125-regression-covers-constants-only.toml

@@ -0,0 +1,16 @@
+id = "1583-rtl8125-regression-covers-constants-only"
+pr = 1583
+title = "修复 realtek-rtl8125：通告千兆自协商能力"
+remote = "https://github.com/rcore-os/tgoskits.git"
+base = "ba27d7b061b47d22b16a53c847b1a4abc2bece71"
+head = "214ccd50c9df852efaa74822341294defa92e001"
+source = "https://github.com/rcore-os/tgoskits/pull/1583#discussion_r3567595139"
+fixed_by = "7b37887e616abdb4b6e98c1e07f533607f51e2b7"
+
+[[expected]]
+id = "rtl8125-regression-does-not-exercise-phy-sequence"
+path = "drivers/net/realtek-rtl8125/src/hw.rs"
+line = 550
+severity = "major"
+description = "该回归测试只检查新常量之间的关系，从未调用 hw_phy_config，也未观察 PHY 读写。因此，即使删除生产代码中的 MII_CTRL1000 和 MII_BMCR 配置，测试仍会通过，原来的自协商缺陷也会复现。"
+match_if = "评审必须指出：新增测试只验证常量，而没有覆盖生产 PHY 配置行为，所以移除新增的页选择、通告和重启序列后测试仍不会失败；并要求通过记录型或 mock PHY 边界，或等价方式建立确定性的生产路径回归测试。"

scripts/agent-review-bench/cases/1611-axvisor-qemu-timeout-watchdog.toml

@@ -0,0 +1,16 @@
+id = "1611-axvisor-qemu-timeout-watchdog"
+pr = 1611
+title = "QEMU：移除 SVM 和 VMX 冒烟测试的显式超时"
+remote = "https://github.com/rcore-os/tgoskits.git"
+base = "dbbe0e06576fd1446c014d4edea3aa480505b5e9"
+head = "6a4a23fe147e02d46c66df1d2fca8269098c2fba"
+source = "https://github.com/rcore-os/tgoskits/pull/1611#discussion_r3583879562"
+fixed_by = "a1a6ed0ce4ac56e006280cca268973578bd7a89d"
+
+[[expected]]
+id = "axvisor-nested-qemu-watchdog-disabled"
+path = "test-suit/axvisor/normal/qemu/smoke/qemu-x86_64-svm.toml"
+line = 25
+severity = "major"
+description = "SVM 与 VMX 嵌套虚拟化冒烟配置删掉了仅有的有限 timeout。解析后 QemuConfig.timeout 变为 None，测试运行器因而不再安装 watchdog；一旦 QEMU 启动、guest shell 或结束条件卡住，任务会无限等待。慢速 runner 已可通过提高基线值或 timeout scale 放宽，不应通过删除超时规避失败。"
+match_if = "评审必须指出：SVM 和 VMX 配置删除 timeout 会使 watchdog 失效，并导致 QEMU 或 guest shell 卡住时无限等待；同时要求为两者恢复有限超时，允许调整基线值或使用 timeout scale。只泛泛建议增加超时或只讨论测试速度不算命中。"

scripts/agent-review-bench/reviewer.md

@@ -0,0 +1,11 @@
+# Offline Review Contract
+
+This file marks the repository as the isolated `review-single-pr offline-benchmark` environment.
+
+- The review target is the committed change between `bench-base` and `HEAD` in this repository.
+- There is no live PR identity or external context. Network access, GitHub operations, paths outside
+  this repository, writes, builds, and tests are unavailable.
+- The current `review-single-pr` skill, `AGENTS.md`, `book/guideline/`, and
+  `.agent-review-context/review.schema.json` are committed on both sides of the synthetic change and
+  therefore are review instructions rather than part of the target diff.
+- The final response must be only the JSON object required by the supplied output schema.

scripts/agent-review-bench/schemas/grade.schema.json

@@ -0,0 +1,26 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "type": "object",
+  "additionalProperties": false,
+  "required": ["matches"],
+  "properties": {
+    "matches": {
+      "type": "array",
+      "items": {
+        "type": "object",
+        "additionalProperties": false,
+        "required": ["expected_id", "finding_index", "reason"],
+        "properties": {
+          "expected_id": { "type": "string" },
+          "finding_index": {
+            "anyOf": [
+              { "type": "integer", "minimum": 0 },
+              { "type": "null" }
+            ]
+          },
+          "reason": { "type": "string" }
+        }
+      }
+    }
+  }
+}

scripts/agent-review-bench/schemas/review.schema.json

@@ -0,0 +1,27 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "type": "object",
+  "additionalProperties": false,
+  "required": ["summary", "findings"],
+  "properties": {
+    "summary": { "type": "string" },
+    "findings": {
+      "type": "array",
+      "items": {
+        "type": "object",
+        "additionalProperties": false,
+        "required": ["title", "body", "path", "line", "severity"],
+        "properties": {
+          "title": { "type": "string" },
+          "body": { "type": "string" },
+          "path": { "type": "string" },
+          "line": { "type": "integer", "minimum": 1 },
+          "severity": {
+            "type": "string",
+            "enum": ["critical", "major", "minor", "nit"]
+          }
+        }
+      }
+    }
+  }
+}