Unified IMA OpenAPI for Notes and Knowledge Bases

What It Solves

When calling the IMA OpenAPI, the hard part is often not the endpoint but encoding, file integrity, and intent routing. If notes writes send GBK, Latin-1, or invalid bytes to https://ima.qq.com, the stored content may become irreversible garbled text. PowerShell 5.1 can also silently convert the request body, turning a successful HTTP call into corrupted data.

How the Skill Works

The skill separates IMA operations into notes and knowledge-base modules, with routing enforced before any API call:
- Notes: search, browse, create, append, and read note content using APIs such as import_doc and append_doc.
- Knowledge base: upload files, add web links, search, browse, and retrieve original content metadata.
- Cross-module flows: requests involving a note inside a knowledge base, or linking a note to a knowledge base, require reading both sub-modules first.

Key steps:
1. Verify credentials from the local config or environment variables.
2. Validate content and title as legal UTF-8 before writing notes.
3. Preserve original file names and binary content during uploads; reject unsupported media such as video links and file:// URLs.
4. Use POST + JSON Body, then inspect both script-level errors in stderr and backend code fields in stdout.

Boundaries and Notes

This is useful for engineers who already have an IMA Client ID and API Key and want agent-driven note and knowledge-base operations. It does not repair invalid encoding, does not normalize uploaded files, and only sends credentials to the official domain. If a -200 update check is returned, the original request has not been sent yet.