Piglet IMA Notes and Knowledge Base

Problem

When an agent needs to operate IMA notes and knowledge bases, failure modes are not only missing API calls, but also wrong module routing, absent credentials, and silent Chinese text corruption from encoding. This skill wraps the ima.qq.com OpenAPI into a decision flow that separates notes from knowledge-base tasks and adds preflight checks for credentials and encoding.

How it works

• Unified auth and transport: uses Client ID and API Key, resolved from environment variables → config file; all requests use HTTP POST + JSON Body and are sent only to the official Base URL.
• Module routing: use notes for note-content actions such as search, browse, read, create, and append; use knowledge-base for knowledge-entry actions such as file upload, web link addition, knowledge search, content browsing, and note association.
• Ambiguity handling: for example, “add content to a note inside a knowledge base” maps to notes append_doc; “add this note to a knowledge base” maps to knowledge-base add_knowledge.
• Encoding protection: notes write paths must validate content and title as valid UTF-8; knowledge-base file uploads keep the original binary content; under PowerShell 5.1, all API request bodies must be explicitly converted to UTF-8 byte arrays.

Boundaries

Useful for teams that already have IMA API credentials and want agent-driven note or knowledge-base workflows. It does not replace the IMA service itself; if credentials are empty, the skill stops and asks for setup. Multi-module tasks run sequentially, for example search from knowledge-base first, then write to notes.