Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when operating Alibaba Cloud through a custom OpenAPI MCP server, especially for SWAS deployment workflows: OAuth setup, selector updates, missing-tool diagnosis, deployment permission closure, and pre-deploy verification.
.claude/skills/ethanyoq-aliyun-openapi-mcp-ops/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 25% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 32% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 28% | 0% |
Category: service
Use this skill when:
This skill is based on a verified SWAS deployment-prep workflow against a custom OpenAPI MCP server.
This skill covers:
This skill does not cover:
OAuth success does not imply deployment selectors are present.
If the id casing is wrong, the browser flow may still open, but the server will not load correctly.
After changing the MCP config or custom server selectors, restart Codex before concluding the tools are still missing.
For deployment preparation, you need remote execution capability.
RunCommandCreateCommandInvokeCommandDeleteCommandCreateSnapshot (recommended before first deploy)If the app uses Feishu long connection and does not need inbound public HTTP immediately, do not add firewall write selectors unless a real deployment step needs them.
Alibaba Cloud Linux 3 is an acceptable clean Linux base for this project.It is suitable for Node + systemd + SQLite + logs + backup scripts.
Keep existing read-only selectors such as:
ListInstancesListInstanceStatusListFirewallRulesListSnapshotsDescribeCloudAssistantStatusDescribeCommandsDescribeInvocationsDescribeInvocationResultAdd deployment selectors:
RunCommandCreateCommandInvokeCommandDeleteCommandCreateSnapshotAdd firewall write selectors only if actually needed later:
CreateFirewallRuleCreateFirewallRulesModifyFirewallRuleEnableFirewallRuleDisableFirewallRuleExample pattern: https://openapi-mcp.<region>.aliyuncs.com/accounts/<account>/custom/<server-name>/id/<server-id>/sse
If the config was just added or changed, restart Codex first.
If only read-only tools exist, do not try to deploy yet.
Use Alibaba Cloud OpenAPI Explorer management APIs to update the custom server.
Do not rely on the current session to pick up new tools.
Runninguname -anode -v || trueSuccess0Before claiming the SWAS side is ready, verify all of the following:
RunningThis skill is valid if it helps you answer these questions quickly and correctly:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 15,738 | 8,548 | -46% | 1 | 1 | 0% | 2,748 | 3,041 | +11% | 0 | 0 | — |
case-02 | fail→fail | 13,933 | 8,432 | -39% | 1 | 1 | 0% | 2,509 | 3,056 | +22% | 0 | 0 | — |
case-03 | pass→pass | 14,795 | 9,052 | -39% | 1 | 1 | 0% | 2,744 | 3,218 | +17% | 0 | 0 | — |
case-04 | pass→pass | 12,122 | 6,083 | -50% | 1 | 1 | 0% | 2,036 | 2,528 | +24% | 0 | 0 | — |
case-05 | fail→pass | 9,478 | 3,316 | -65% | 1 | 1 | 0% | 1,602 | 2,010 | +25% | 0 | 0 | — |
case-06 | fail→pass | 11,845 | 4,593 | -61% | 1 | 1 | 0% | 1,905 | 2,096 | +10% | 0 | 0 | — |
case-07 | fail→pass | 10,122 | 5,654 | -44% | 1 | 1 | 0% | 1,813 | 2,386 | +32% | 0 | 0 | — |
case-08 | fail→pass | 12,585 | 6,343 | -50% | 1 | 1 | 0% | 1,958 | 2,507 | +28% | 0 | 0 | — |
case-09 | pass→pass | 11,348 | 3,236 | -71% | 1 | 1 | 0% | 1,983 | 2,058 | +4% | 0 | 0 | — |
case-10 | fail→pass | 9,113 | 2,633 | -71% | 1 | 1 | 0% | 1,614 | 1,859 | +15% | 0 | 0 | — |
case-15 | pass→pass | 6,604 | 3,876 | -41% | 1 | 1 | 0% | 1,142 | 2,153 | +89% | 0 | 0 | — |
case-11 | pass→pass | 9,368 | 4,581 | -51% | 1 | 1 | 0% | 1,610 | 2,267 | +41% | 0 | 0 | — |
case-12 | fail→pass | 9,085 | 3,887 | -57% | 1 | 1 | 0% | 1,465 | 1,983 | +35% | 0 | 0 | — |
case-13 | pass→pass | 10,528 | 4,386 | -58% | 1 | 1 | 0% | 1,847 | 2,205 | +19% | 0 | 0 | — |
case-14 | fail→pass | 12,473 | 3,460 | -72% | 1 | 1 | 0% | 2,262 | 2,102 | -7% | 0 | 0 | — |
case-16 | pass→pass | 7,751 | 2,690 | -65% | 1 | 1 | 0% | 1,217 | 1,899 | +56% | 0 | 0 | — |
case-17 | pass→pass | 11,741 | 2,396 | -80% | 1 | 1 | 0% | 2,057 | 1,888 | -8% | 0 | 0 | — |
case-18 | pass→pass | 10,426 | 3,941 | -62% | 1 | 1 | 0% | 1,877 | 2,170 | +16% | 0 | 0 | — |
case-19 | fail→pass | 11,866 | 5,870 | -51% | 1 | 1 | 0% | 2,155 | 2,552 | +18% | 0 | 0 | — |
case-20 | pass→pass | 11,683 | 9,766 | -16% | 1 | 1 | 0% | 2,335 | 3,379 | +45% | 0 | 0 | — |
case-21 | pass→pass | 8,999 | 8,063 | -10% | 1 | 1 | 0% | 1,891 | 3,124 | +65% | 0 | 0 | — |
case-22 | pass→pass | 3,661 | 3,536 | -3% | 1 | 1 | 0% | 634 | 1,971 | +211% | 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 +41 percentage points is the difference between those two pass rates over the 22 comparable cases.
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.