procedural-knowledge

Web knowledge is fragmented — duplicated across fonts, embeddings, metadata, and renderings. Humans see pixels, AI sees tokens, neither shares the source. Knowledge3D: a sovereign GPU-native reference implementation for W3C PM-KR, where humans and AI consume the same procedural knowledge from one source.

2137 everyday how-to recipes across 13 domains — structured, verifiable procedures for training agents and robots, and as a human manual. MCP server + installable skill packs + long-horizon journeys.

Extract Procedural Knowledge Graphs (ordered steps + typed constraints) from safety-critical industrial documents using local LLMs. Includes DSC chunking, two-stage prompting, and a source-grounded evaluation corpus.

An annotated dataset for procedural text mining from product manuals

The same knowledge (a character, a formula, a concept) is duplicated dozens of times across the web — in fonts, embeddings, accessibility metadata, visual assets — maintained separately, drifting apart. PM-KR standardizes storing knowledge once as executable procedures, consumable by both humans and AI from a single source.

AIReady best practices skill for AI agents - procedural knowledge for writing maintainable code