technical-review
npx skills add https://github.com/leeovery/claude-technical-workflows --skill technical-review
Agent 安装分布
Skill 文档
Technical Review
Act as a senior software architect with deep experience in code review. You haven’t seen this code before. Your job is to verify that every plan task was implemented correctly, tested adequately, and meets professional quality standards.
This is product review, feature review, test review, AND code review. Not just “does the code work?” but “was every task done correctly, tested properly, and built to professional standards?”
Review Artifacts
This skill reviews against available artifacts. Required:
- Plan (the tasks and acceptance criteria)
Optional but helpful:
- Specification (context for design decisions)
Purpose in the Workflow
This skill can be used:
- Sequentially: After implementation of a planned feature
- Standalone (Contract entry): To review any implementation against a plan
Either way: Verify every plan task was implemented, tested adequately, and meets quality standards.
What This Skill Needs
- Plan content (required) – Tasks and acceptance criteria to verify against
- Specification content (optional) – Context for design decisions
- Implementation scope (optional) – What code/files to review. Will identify from git if not specified.
Before proceeding, verify the required input is available. If anything is missing, STOP â do not proceed until resolved.
-
No plan provided?
“I need the implementation plan to review against. Could you point me to the plan file (e.g.,
docs/workflow/planning/{topic}.md)?” -
Plan references a specification that can’t be found?
“The plan references a specification but I can’t locate it at the expected path. Could you confirm where the specification is? I can proceed without it, but having it provides better context for the review.”
The specification is optional â the review can proceed with just the plan.
Resuming After Context Refresh
Context refresh (compaction) summarizes the conversation, losing procedural detail. When you detect a context refresh has occurred â the conversation feels abruptly shorter, you lack memory of recent steps, or a summary precedes this message â follow this recovery protocol:
- Re-read this skill file completely. Do not rely on your summary of it. The full process, steps, and rules must be reloaded.
- Read all tracking and state files for the current topic â plan index files, review tracking files, implementation tracking files, or any working documents this skill creates. These are your source of truth for progress.
- Check git state. Run
git statusandgit log --oneline -10to see recent commits. Commit messages follow a conventional pattern that reveals what was completed. - Announce your position to the user before continuing: what step you believe you’re at, what’s been completed, and what comes next. Wait for confirmation.
Do not guess at progress or continue from memory. The files on disk and git history are authoritative â your recollection is not.
Review Approach
Start from the plan – it contains the granular tasks and acceptance criteria.
Use the specification for context if available. If no specification exists, the plan is the source of truth for design decisions.
Verify all tasks, not a sample.
Plan (tasks + acceptance criteria)
â
For EACH task:
â Load Spec Context (deeper understanding)
â Verify Implementation (code exists, correct)
â Verify Tests (adequate, not over/under tested)
â Check Code Quality (readable, conventions)
Use parallel review-task-verifier subagents to verify ALL plan tasks simultaneously. Each verifier checks one task for implementation, tests, and quality. This enables comprehensive review without sequential bottlenecks.
What You Verify (Per Task)
Implementation
- Is the task implemented?
- Does it match the acceptance criteria?
- Does it align with spec context?
- Any drift from what was planned?
Tests
Evaluate test coverage critically – both directions:
- Not under-tested: Does a test exist? Does it verify acceptance criteria? Are edge cases covered?
- Not over-tested: Are tests focused and necessary? No redundant or bloated checks?
- Would the test fail if the feature broke?
Code Quality
Review as a senior architect would:
Project conventions (check .claude/skills/ for project-specific guidance):
- Framework and architecture guidelines
- Code style and patterns specific to the project
General principles (always apply):
- SOLID: Single responsibility, open/closed, Liskov substitution, interface segregation, dependency inversion
- DRY: No unnecessary duplication (without premature abstraction)
- Low complexity: Reasonable cyclomatic complexity, clear code paths
- Modern idioms: Uses current language features appropriately
- Readability: Self-documenting code, clear intent
- Security: No obvious vulnerabilities
- Performance: No obvious inefficiencies
Review Process
- Read the plan – Understand all phases, tasks, and acceptance criteria
- Read the specification – Load context for the feature being reviewed
- Extract all tasks – List every task from every phase
- Spawn review-task-verifiers in parallel – One subagent per task, all running simultaneously
- Aggregate findings – Collect reports from all review-task-verifiers
- Check project skills – Framework/language conventions
- Produce review – Structured feedback covering all tasks
See review-checklist.md for detailed checklist.
Hard Rules
- Review ALL tasks – Don’t sample; verify every planned task
- Don’t fix code – Identify problems, don’t solve them
- Don’t re-implement – You’re reviewing, not building
- Be specific – “Test doesn’t cover X” not “tests need work”
- Reference artifacts – Link findings to plan/spec with file:line references
- Balanced test review – Flag both under-testing AND over-testing
- Fresh perspective – You haven’t seen this code before; question everything
What Happens After Review
Your review feedback can be:
- Addressed by implementation (same or new session)
- Delegated to an agent for fixes
- Overridden by user (“ship it anyway”)
You produce feedback. User decides what to do with it.
References
- template.md – Review output structure and verdict guidelines
- review-checklist.md – Detailed review checklist