Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Manage Anthropic workspaces, API keys, team access, and spending limits Use when working with enterprise-rbac patterns. for enterprise Claude deployments. Trigger with "anthropic workspace", "anthropic team management", "claude enterprise", "anthropic api key management".
.claude/skills/jeremylongshore-clade-enterprise-rbac/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | -11% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-16 | ✓→✗ | ▼ Worse | -5% | 0% |
| case-11 | ✓→✓ | ＝ Same ✓ | 18% | 0% |
Anthropic uses Organizations and Workspaces for access control. API keys are scoped to workspaces.
Organization (your-company)
├── Workspace: Production
│ ├── API Key: prod-backend (Tier 4)
│ └── API Key: prod-frontend-proxy (Tier 2)
├── Workspace: Staging
│ └── API Key: staging-all (Tier 2)
└── Workspace: Development
└── API Key: dev-team (Tier 1)| Practice | Why | |----------|-----| | One key per service/environment | Isolate blast radius | | Name keys descriptively | prod-recommendation-service not key-1 | | Set spending limits per key | Prevent runaway costs from bugs | | Rotate quarterly | Reduce exposure window | | Never share dev and prod keys | Different rate limit tiers |
Set in Anthropic Console → Settings → Limits:
| Error | Cause | Solution | |-------|-------|----------| | API Error | Check error type and status code | See clade-common-errors |
See Organization Structure diagram, API Key Best Practices table, and Access Control Checklist above.
See clade-migration-deep-dive for migrating from other LLM providers.
Each section contains production-ready code examples. Copy and adapt them to your use case.
Integrate the patterns that match your requirements. Test each change individually.
Run your test suite to confirm the integration works correctly.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-11 | pass→pass | 5,508 | 2,989 | -46% | 1 | 1 | 0% | 706 | 831 | +18% | 0 | 0 | — |
case-01 | pass→pass | 39,553 | 11,377 | -71% | 1 | 1 | 0% | 3,297 | 2,473 | -25% | 0 | 0 | — |
case-02 | pass→pass | 14,822 | 5,465 | -63% | 1 | 1 | 0% | 2,292 | 1,611 | -30% | 0 | 0 | — |
case-03 | pass→pass | 13,276 | 12,459 | -6% | 1 | 1 | 0% | 2,066 | 2,581 | +25% | 0 | 0 | — |
case-04 | pass→pass | 9,602 | 4,644 | -52% | 1 | 1 | 0% | 1,566 | 1,397 | -11% | 0 | 0 | — |
case-05 | pass→pass | 13,532 | 5,383 | -60% | 1 | 1 | 0% | 2,144 | 1,555 | -27% | 0 | 0 | — |
case-06 | pass→pass | 11,644 | 5,292 | -55% | 1 | 1 | 0% | 1,780 | 1,600 | -10% | 0 | 0 | — |
case-07 | pass→pass | 11,447 | 5,470 | -52% | 1 | 1 | 0% | 1,749 | 1,404 | -20% | 0 | 0 | — |
case-08 | pass→pass | 7,577 | 4,601 | -39% | 1 | 1 | 0% | 1,261 | 1,403 | +11% | 0 | 0 | — |
case-09 | fail→pass | 7,170 | 2,178 | -70% | 1 | 1 | 0% | 982 | 873 | -11% | 0 | 0 | — |
case-10 | pass→pass | 7,417 | 3,119 | -58% | 1 | 1 | 0% | 1,103 | 1,080 | -2% | 0 | 0 | — |
case-12 | pass→pass | 11,638 | 10,088 | -13% | 1 | 1 | 0% | 1,866 | 1,534 | -18% | 0 | 0 | — |
case-13 | pass→pass | 3,769 | 1,572 | -58% | 1 | 1 | 0% | 548 | 828 | +51% | 0 | 0 | — |
case-14 | pass→pass | 8,532 | 3,829 | -55% | 1 | 1 | 0% | 1,294 | 1,227 | -5% | 0 | 0 | — |
case-15 | pass→pass | 13,460 | 6,529 | -51% | 1 | 1 | 0% | 1,900 | 1,610 | -15% | 0 | 0 | — |
case-16 | pass→fail | 7,759 | 2,396 | -69% | 1 | 1 | 0% | 1,043 | 989 | -5% | 0 | 0 | — |
case-17 | fail→pass | 11,223 | 2,644 | -76% | 1 | 1 | 0% | 1,789 | 1,024 | -43% | 0 | 0 | — |
case-18 | fail→pass | 5,750 | 1,597 | -72% | 1 | 1 | 0% | 901 | 861 | -4% | 0 | 0 | — |
case-19 | pass→pass | 1,879 | 1,916 | +2% | 1 | 1 | 0% | 181 | 853 | +371% | 0 | 0 | — |
case-20 | pass→pass | 12,717 | 12,497 | -2% | 1 | 1 | 0% | 2,666 | 3,209 | +20% | 0 | 0 | — |
case-21 | pass→pass | 12,781 | 9,829 | -23% | 1 | 1 | 0% | 2,485 | 2,587 | +4% | 0 | 0 | — |
case-22 | pass→pass | 16,899 | 17,675 | +5% | 1 | 1 | 0% | 2,868 | 3,588 | +25% | 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. The headline lift of +9 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is 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.