Product background, now building AI applications. Working on knowledge-management tooling and agent infrastructure.
- Consolidating skills scattered across tool directories into a single source of truth
- Restructuring an Obsidian vault into an agent-retrievable knowledge base
- TypeScript for frontends and CLIs, Python for scripts and data pipelines
| Focus | What I'm doing |
|---|---|
| Agent skill library | Building a local skill library — 180+ skills consolidated into one source of truth, distributed to multiple CLI clients via junctions |
| Knowledge-base engineering | AI-first refactor of an Obsidian vault: structured frontmatter, a traversable wikilink graph, and agent-facing retrieval protocols |
| AI application development | Learning by building. TypeScript for frontends and CLIs, Python for scripts and data pipelines |
Skill consolidation Merging skills scattered across tool directories into a single source of truth. The hard part isn't copying files — it's that Windows junctions are transparent: any tool that clears its own skills directory wipes the source too.
The knowledge base as agent context Notes aren't written for people to read — they're written for future agents to retrieve. Each note carries its own summary, sources, and confidence level, so an agent with zero context can judge relevance in 10 seconds.
Validating from zero No client work, no capability arms race. Build something priceable first, then talk about the rest.
Profile README lives in NixRonox/NixRonox
