Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Full-repo SAP CC Go compliance audit against review standards.
.claude/skills/notque-sapcc-audit/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -37% | 0% |
Review every package against established review standards. Not checklist compliance — code-level review that finds over-engineering, dead code, interface violations, and inconsistent patterns.
| Signal | Load This File | When | |--------|---------------|------| | Phase 1 begins | references/phase-1-discover-commands.md | Detection commands, package mapping, segmentation table | | Phase 2 begins | references/phase-2-dispatch-agents.md | Full dispatch prompt and per-domain review checklist (11 areas) | | Phase 3 begins | references/output-templates.md | Report scaffold, per-finding format, severity guide |
Load each reference file at the start of its phase. Do not load all three upfront.
Goal: Map the repository and plan the package segmentation.
Read references/phase-1-discover-commands.md for the exact detection commands, segmentation table, and file-count queries.
Verify this is an sapcc project (sapcc imports in go.mod). If not, stop immediately.
Map all packages, count files per package, and produce a segmentation table (5–8 agents, 5–15 files each).
Gate: Packages mapped, agents planned. Proceed to Phase 2.
Goal: Launch parallel agents that review packages against project standards.
Read references/phase-2-dispatch-agents.md for the full dispatch prompt (11 review areas: over-engineering, dead code, error messages, constructors, interface contracts, copy-paste, HTTP handlers, database patterns, type patterns, logging, mixed approaches).
Use the standard dispatch prompt verbatim, substituting the assigned package list.
Dispatch all agents in a single message using the Task tool with subagent_type=golang-general-engineer.
Gate: All agents dispatched. Proceed to Phase 3.
Goal: Aggregate findings into a code-level compliance report.
Read references/output-templates.md for the report scaffold, per-finding format, and deduplication rules.
Deduplicate by file:line. Write sapcc-audit-report.md. Display verdict, must-fix count, and top 5 findings inline.
Gate: Report complete.
| Scenario | Response | |----------|----------| | Not an sapcc project | Stop immediately. Print: "This does not appear to be an SAP CC Go project (no sapcc imports in go.mod)." | | Agents cannot read a file | Log and continue. Flag in the report under "Warnings." | | gopls MCP tools unavailable | Fall back to manual grep-based analysis. Note in the report. | | Too many packages (>30) | Split into >8 agents. Ensure each still gets 5-15 files. | | Agent finds no violations | Report is valid. Output empty sections for unused severity levels. |
Audit only: READS and REPORTS. Does NOT modify code unless explicitly asked with --fix.
/do routes via "sapcc audit", "sapcc compliance", "sapcc lead review"go-patterns (the rules), golang-general-engineer (the executor)| Package Type | Reference to Load | |-------------|-------------------| | HTTP handlers (internal/api/) | api-design-detailed.md | | Test files (*_test.go) | testing-patterns-detailed.md | | Error handling heavy packages | error-handling-detailed.md | | Architecture/drivers | architecture-patterns.md | | Build/CI config | build-ci-detailed.md | | Import-heavy files | library-reference.md |
Always available for calibration (load only when needed): quality-issues.md, review-standards-lead.md.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | fail→pass | 37,268 | 12,413 | -67% | 1 | 1 | 0% | 3,017 | 2,442 | -19% | 0 | 0 | — |
case-01 | fail→fail | 5,126 | 36,225 | +607% | 1 | 1 | 0% | 744 | 1,192 | +60% | 0 | 0 | — |
case-02 | fail→fail | 29,129 | 5,153 | -82% | 1 | 1 | 0% | 5,030 | 1,091 | -78% | 0 | 0 | — |
case-03 | fail→fail | 30,486 | 4,344 | -86% | 1 | 1 | 0% | 4,937 | 1,115 | -77% | 0 | 0 | — |
case-04 | fail→fail | 20,409 | 5,183 | -75% | 1 | 1 | 0% | 3,336 | 1,147 | -66% | 0 | 0 | — |
case-06 | pass→pass | 11,303 | 3,843 | -66% | 1 | 1 | 0% | 1,534 | 1,241 | -19% | 0 | 0 | — |
case-07 | fail→fail | 12,331 | 2,361 | -81% | 1 | 1 | 0% | 1,693 | 1,218 | -28% | 0 | 0 | — |
case-08 | fail→pass | 9,146 | 1,947 | -79% | 1 | 1 | 0% | 1,394 | 1,191 | -15% | 0 | 0 | — |
case-09 | fail→pass | 9,500 | 2,458 | -74% | 1 | 1 | 0% | 1,344 | 1,230 | -8% | 0 | 0 | — |
case-10 | pass→pass | 15,829 | 3,218 | -80% | 1 | 1 | 0% | 2,392 | 1,465 | -39% | 0 | 0 | — |
case-11 | fail→pass | 12,689 | 1,928 | -85% | 1 | 1 | 0% | 2,041 | 1,155 | -43% | 0 | 0 | — |
case-12 | fail→pass | 12,526 | 2,106 | -83% | 1 | 1 | 0% | 1,954 | 1,236 | -37% | 0 | 0 | — |
case-13 | fail→pass | 12,894 | 2,603 | -80% | 1 | 1 | 0% | 1,782 | 1,410 | -21% | 0 | 0 | — |
case-14 | fail→fail | 11,069 | 2,298 | -79% | 1 | 1 | 0% | 1,550 | 1,224 | -21% | 0 | 0 | — |
case-15 | fail→fail | 14,436 | 1,972 | -86% | 1 | 1 | 0% | 1,060 | 1,225 | +16% | 0 | 0 | — |
case-16 | fail→fail | 12,817 | 3,112 | -76% | 1 | 1 | 0% | 1,873 | 1,424 | -24% | 0 | 0 | — |
case-17 | fail→fail | 13,983 | 3,680 | -74% | 1 | 1 | 0% | 2,005 | 1,436 | -28% | 0 | 0 | — |
case-18 | pass→pass | 14,326 | 3,772 | -74% | 1 | 1 | 0% | 2,106 | 1,462 | -31% | 0 | 0 | — |
case-19 | fail→pass | 11,793 | 3,936 | -67% | 1 | 1 | 0% | 1,661 | 1,519 | -9% | 0 | 0 | — |
case-20 | fail→fail | 7,819 | 10,579 | +35% | 1 | 1 | 0% | 1,381 | 1,216 | -12% | 0 | 0 | — |
case-21 | pass→fail | 16,855 | 6,570 | -61% | 1 | 1 | 0% | 3,422 | 1,339 | -61% | 0 | 0 | — |
case-22 | pass→fail | 4,039 | 7,673 | +90% | 1 | 1 | 0% | 527 | 1,211 | +130% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 15 counted toward the lift figure. The other 7 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +23 percentage points is the difference between those two pass rates over the 15 comparable cases. 4 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.