基于 SOC 职业分类
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/notque/vexjoy-agent --skill sapcc-review命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
Run the full evidence-to-live implementation workflow for large, multi-system, multi-wave, or CPU-delegated 5 Star Booker GM programs.
Classify user requests and route to the correct agent + skill. Primary entry point for all delegated work.
Structured multi-phase workflows: review, debug, refactor (tidy, clean up, untangle messy code without behaviour change), deploy, create, research.
| name | sapcc-review |
| description | Gold-standard SAP CC Go code review: 10 parallel domain specialists. |
| user-invocable | false |
| argument-hint | [--fix] |
| command | /sapcc-review |
| agent | golang-general-engineer |
| allowed-tools | ["Agent","Read","Write","Bash","Grep","Glob","Edit","TaskCreate","TaskUpdate","TaskList","EnterWorktree"] |
| routing | {"not_for":"full-repo compliance audit scan (use sapcc-audit). This skill is deep 10-specialist parallel review.","triggers":["sapcc review","sapcc lead review","sapcc compliance review","full sapcc check","review sapcc standards"],"pairs_with":["golang-general-engineer","go-patterns","sapcc-audit"],"force_route":false,"complexity":"Complex","category":"language"} |
10-agent domain-specialist review. Each agent masters one rule domain and scans every package for violations against the comprehensive patterns reference.
How this differs from /sapcc-audit: sapcc-audit segments by package (generalist per package). sapcc-review segments by rule domain (specialist per concern, cross-package). Both are useful; this one catches more because specialists find issues generalists miss.
This skill executes a gold-standard code review against SAP Converged Cloud Go repository standards through parallel domain specialists. Rather than one generalist reviewing one package, ten specialists review all packages for their specific domain (error handling, testing, types, HTTP APIs, etc.). This catches systemic patterns that package-level reviews miss.
Each specialist loads only its domain-specific reference file to keep context tight and focus deep. Findings are code-level (actual rejected/correct examples, never abstract suggestions) and cite specific sections from sapcc-code-patterns.md.
| Signal | Load These Files | Why |
|---|---|---|
| Phase 2 DISPATCH: the 10 domain-specialist prompts | agent-dispatch-prompts.md | Loads detailed guidance from agent-dispatch-prompts.md. |
| Phase 3 AGGREGATE: report structure and severity boosts | report-template.md | Loads detailed guidance from report-template.md. |
Goal: Map the repository, verify it's an sapcc project, plan the review.
Step 1: Verify sapcc project
cat go.mod | head -5
grep -c "sapcc" go.mod
If the module path doesn't contain "sapcc" AND go.mod doesn't import any sapcc packages, warn the user but continue (they may want to check a non-sapcc repo against the project's standards).
Step 2: Map all Go packages and files
# Count .go files (excluding vendor)
find . -name "*.go" -not -path "*/vendor/*" | wc -l
find . -name -not -path | sed | | -c | -rn
find . -name -not -path | -l
Step 3: Check for key imports (determines which rules are most relevant)
grep -r "go-bits" go.mod # Uses go-bits?
grep -r "go-api-declarations" go.mod # Uses API declarations?
grep -r "gophercloud" go.mod # Uses OpenStack?
grep -r "gorilla/mux" go.mod # HTTP routing?
grep -r "database/sql" go.mod # Database?
Step 4: Create task_plan.md
# Task Plan: SAPCC Review — [repo name]
## Goal
Comprehensive code review of [repo] against project standards, dispatching 10 domain-specialist agents.
## Phases
- [x] Phase 1: Discover repo structure
- [ ] Phase 2: Dispatch 10 specialist agents
- [ ] Phase 3: Aggregate findings
- [ ] Phase 4: Write report
## Repo Profile
- Module: [module path]
- Packages: [N]
- Go files: [M] (excluding vendor)
- Test files: [T]
- Key imports: [list]
## Status
**Currently in Phase 2** - Dispatching agents
Gate: Repo mapped, plan created. Proceed to Phase 2.
Goal: Launch 10 domain-specialist agents in a SINGLE message for true parallel execution.
CRITICAL: All 10 agents must be dispatched in ONE message using the Agent tool. Dispatch them in one message. Parallel domain specialists work independently on disjoint concerns, so a single dispatch keeps the review moving.
Each agent receives:
See references/agent-dispatch-prompts.md for the shared preamble and all 10 agent specifications (Agents 1-10).
Gate: All 10 agents dispatched in single message. Wait for all to complete. Proceed to Phase 3.
Goal: Compile all agent findings into a single prioritized report.
Step 0: Full file inventory
Run git status --short (not just git diff --stat) to capture both modified AND untracked (new) files. This ensures new files created during the review session are not missed in the report.
Step 1: Collect all findings
Read each agent's output. Extract all findings with their severity, file, rule, and code.
Step 2: Deduplicate
If two agents flagged the same file:line, keep the higher-severity finding with the more specific rule citation.
Step 3: Prioritize
Apply cross-repository reinforcement from §35. See references/report-template.md for the severity boost table.
Step 4: Identify Quick Wins
Mark findings that are:
These go in a "Quick Wins" section at the top of the report.
Step 5: Write report
Create sapcc-review-report.md using the full template in references/report-template.md.
Gate: Report written. Display summary to user. Proceed to Phase 4 if --fix specified.
--fix flag)Goal: Apply fixes on an isolated branch.
Step 1: Create worktree
Use EnterWorktree to create an isolated copy. Name it sapcc-review-fixes.
Step 2: Apply Quick Wins first
Start with Quick Wins (lowest risk). After each group of fixes:
go build ./... # Must still compile
go vet ./... # Must pass vet
make check 2>/dev/null || go test ./... # Must pass tests
Step 3: Apply Critical and High fixes
Apply in order. Run tests between each fix. If a fix breaks tests, revert it and note in the report.
Step 4: Create commit
git add -A
git commit -m "fix: apply sapcc-review findings (N fixes across M files)"
Step 5: Report results
Update sapcc-review-report.md with:
When an agent fails or produces empty findings:
When a finding looks wrong (e.g., false positive):
When --fix breaks tests:
review-standards-lead.mdarchitecture-patterns.mdapi-design-detailed.mderror-handling-detailed.mdtesting-patterns-detailed.mdarchitecture-patterns.mdpreferred-patterns.mdpr-mining-insights.md — Review severity calibration across projectslibrary-reference.md — Approved/forbidden dependency tablereferences/agent-dispatch-prompts.md — Shared preamble + all 10 agent specificationsreferences/report-template.md — Full report template + severity boost tableIntegration notes:
/sapcc-audit (package-level generalist) — use both for maximum coverage~/.claude/skills/engineering/go-patterns/cp -r skills/engineering/sapcc-review ~/.claude/skills/sapcc-review for global access/do routes via "sapcc review", "sapcc lead review", "comprehensive sapcc audit"