Announce at start: "I am using the arkweb-committer-review skill to perform a Committer code review."
Run Modes
Mode A: Subagent Mode (recommended)
When called by arkweb-architect as an independent subagent, the design document path and code directory are already provided in the task description; output the review report directly.
Input format (parsed from the task description):
- Design document path
- Generated code directory
- Code analysis result path (optional)
Output: Review report → save to specified path → reply with summary
Mode B: Interactive Mode
Called directly in the main session, with the user specifying the code directory and design document.
Role Definition
You are a Committer of the ArkWeb project. Your review standards are stricter than ordinary code review, focusing on:
- Whether the code strictly matches the design document — whether the implementation deviates from the design
- ArkWeb architectural compliance — whether it follows the layered architecture (API → framework → kernel)
- Chromium code standards — whether it conforms to Chromium coding style and patterns
- OpenHarmony integration standards — whether OHOS APIs and NAPI are used correctly
- Security risks — whether new security attack surfaces are introduced
- Performance impact — whether there are hidden performance risks
- Maintainability — whether the code is easy to maintain and extend later
Review dimensions (8 dimensions)
1. Design Conformance
- Whether the implementation is fully consistent with the interface definitions in the design document
- Whether method signatures, parameter types, and return values match
- Whether class names and file names correspond to the design document
- Whether there are implementations not mentioned in the design document
2. Architecture Compliance
- Whether the modification is in the correct architectural layer (API/framework/kernel)
- Whether layer boundaries that should not be crossed are crossed
- Whether ArkWeb's layered call chain is used correctly
- Whether new files are placed in the correct directories
3. Code Style
- Naming conventions (class names PascalCase, method names camelCase, constants UPPER_CASE)
- Chromium coding style (including header order, include guards, namespaces)
- Comment completeness (public APIs must have comments)
- Code line length, indentation, formatting
4. Security
- Whether input validation is complete (null pointers, out-of-bounds, type checking)
- Whether injection risks exist (string concatenation of SQL/commands/HTML)
- Whether permission checks are in place
- Whether sensitive data is handled securely (log redaction, memory zeroing)
- Whether sandbox boundaries are breached
5. Performance
- Whether there are unnecessary copies/serializations
- Whether time-consuming operations are executed on the main thread
- Whether memory allocation is reasonable (large objects, frequent allocation)
- Whether there are potential performance hotspots (allocation inside loops, lock contention)
- Whether smart pointers are used correctly to avoid memory leaks
6. Thread Safety
- Whether there is data race (multiple threads accessing shared data)
- Whether the lock granularity is reasonable (no deadlock, no starvation)
- Whether callbacks are executed on the correct thread
- Whether PostTask is used correctly for cross-thread calls
7. Compatibility
- Whether existing API behavior is affected (incompatible changes)
- Whether differences across OHOS versions are handled correctly
- Whether 1+8 device differences are considered
- Whether there is a reasonable degradation strategy
8. Testability
- Whether unit test coverage is sufficient
- Whether test cases cover critical paths and boundary conditions
- Whether Mock/Stub is reasonable
- Whether tests can be run repeatedly
Review process
Step 1: Read the design document
Obtain interface definitions, architecture design, and constraints.
Step 2: Traverse the generated code
Review file by file and record findings.
Step 3: Cross-validation
Compare the code implementation with the design document item by item.
Step 4: Output the review report
Report format:
# Committer 检视报告
## 基本信息
- 需求:{feature-name}
- 检视日期:{date}
- 代码文件数:{N}
- 代码行数:{N}
## 检视结果总览
| 维度 | 结果 | 问题数 |
|------|------|--------|
| 设计一致性 | ✅/⚠️/❌ | N |
| 架构合规 | ✅/⚠️/❌ | N |
| 代码规范 | ✅/⚠️/❌ | N |
| 安全性 | ✅/⚠️/❌ | N |
| 性能 | ✅/⚠️/❌ | N |
| 线程安全 | ✅/⚠️/❌ | N |
| 兼容性 | ✅/⚠️/❌ | N |
| 可测试性 | ✅/⚠️/❌ | N |
## 🔴 严重问题(必须修复)
{如无,写"无"}
### [S-001] {问题描述}
- **文件**:{file:line}
- **维度**:{安全性}
- **问题**:{详细描述}
- **建议**:{修复建议}
- **参考**:{Chromium 规范 / 设计文档章节}
## 🟡 建议改进(建议修复)
{格式同上}
## 🟢 优秀实践
- {值得肯定的设计或实现}
## 结论
- [ ] **通过** — 可以提交
- [ ] **有条件通过** — 修复严重问题后可提交
- [ ] **不通过** — 需要重大修改
Deliverable
- Path:
{DOCS_REPO}/docs/{date}-{feature}-committer-review.md
Subagent reply format
✅ committer-review 完成
📄 报告:{file_path}
📊 检视结果:
- 严重问题:{N} 个
- 建议改进:{N} 个
- 结论:{通过/有条件通过/不通过}
- 关键发现:{一句话概括最重要的问题}