Pontx Hub

Pontx Hub

Use pontx-hub as the authoritative interface to the curated API catalog. Keep API metadata out of long-lived context by searching and loading only the product, endpoint, or schema needed for the current task.

This universal Skill owns catalog-wide discovery and request safety. A pontx- product Skill adds only provider-specific integration flows, best practices, and caveats; it does not replace the current PontxSpec.

Workflow

1. Search before choosing an API. Natural-language capability, request, and response questions are supported:

   pontx-hub search "" --json
   pontx-hub search "创建任务的入参" --locale zh --json
   pontx-hub search "返回 dueDate 的接口" --locale zh --json
   

   Narrow the result type when useful:

   pontx-hub search "" --type schema --json
   

2. Inspect the selected stable resource ID:

   pontx-hub show endpoint:/
   pontx-hub show schema:/
   

3. When provider-specific guidance would help, check the available product Skills. Explain the benefit and install one only when the user requests or approves the installation:

   pontx-hub skill list
   pontx-hub skill install 
   

   Use that product Skill for provider-specific sequencing and caveats, then return to pontx-hub show for current Endpoint, Schema, and auth details.

4. For an Endpoint, build and review the exact request without sending it:

   pontx-hub  preview  --parameter value --body ''
   pontx-hub  preview   --parameter value --body ''
   

5. For GET or HEAD, call only when the user requested execution:

   pontx-hub  call  --parameter value
   pontx-hub  call   --parameter value
   

6. For POST, PUT, PATCH, or DELETE, show the dry-run result and obtain explicit user confirmation. Then pass --yes without changing parameters:

   pontx-hub  call   --parameter value --body '' --yes
   

7. Generate application code after the request shape is verified. Use the package and exports reported by the Hub instead of inventing a package or pinning a stale version:

   pontx-hub sdk 
   pnpm add @pontx/
   

Safety

• Never infer approval for a write from a request to inspect, explain, or preview an API.
• Never print, echo, log, or place credentials directly in command arguments.
• Read credentials from the environment variables named by show.
• Never call an arbitrary URL or an Endpoint outside the Hub catalog.
• Never treat a product Skill as the source for Endpoint, parameter, Schema, authentication, package-version, or server metadata; inspect the current PontxSpec through the Hub.
• Never bypass the separate preview step for a mutation.
• Treat a changed parameter, body, server, or Endpoint as a new request that requires another dry-run and confirmation.
• Read references/auth-and-safety.md when configuring credentials, diagnosing authentication, or executing a mutation.

Output

Prefer --json when consuming CLI output programmatically. Preserve the CLI exit code and machine-readable error code when reporting failures. Use the result's match.mode and match.fields to explain whether product, parameter, request, response, schema, or property metadata produced the match. Use the Endpoint's published @pontx/ package when generating application code.