Look at the example below and you can see what looks like a Notenik file. Having all of the front matter fields defined in a collection, with their options set as choices, seems like it would be a great tool for generating agentic tasks.
More control over subagents
You get a lot of control over a custom subagent from its frontmatter. You can pick its model and effort, limit its tools, cap its turns, give it its own worktree or memory
.claude/agents/log-scanner.md
---
name: log-scanner
description: Scans CI logs for the first failing step and reports it
model: sonnet
effort: low
memory: project
omitClaudeMd: true
---
Find the first failing step in the log you are given and report the command, the error and the file it points to.
I don’t use agents but I don’t see why an agent wouldn’t be able to access an entire Notenik collection formatted like this. You could probably dictate an entire workflow from a collection.
This is something I’ve been skeptic of. But after reading Andrej Karpathy’s explanation (ex-OpenAI/current Anthropic) in limited cases I can see how it can be useful.
TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it’s the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.
Between the periphrastic output and my waning interest in “chatting” with Claude an “LLM wiki” could be a decent alternative frontend.
