Keeping knowledge organized
A growing Markdown vault becomes hard to use when capture is faster than organization. Mana’s job is to turn each useful input into maintained knowledge: source material, a readable page, links to related pages, and explicit caveats where the record is uncertain.
Files have roles
Section titled “Files have roles”Mana’s default wiki layout separates the work into a few plain-file layers:
raw/keeps source material and provenance.sources/contains readable overviews of individual raw sources.synthesis/contains pages that connect multiple sources into decisions, summaries, checklists, or explanations.
This keeps the repo from becoming a pile of undifferentiated notes. Raw material stays available, but readers usually navigate through source overviews and synthesis pages.
Ground claims
Section titled “Ground claims”Ask Mana to ground claims. The agent searches the web, saves the sources it uses, and adds provenance for important facts so the page can be checked later.
prompt
ground claims. Search the web and back important facts with two reputable sources where possible.
Resolve contradictions
Section titled “Resolve contradictions”Ask Mana to resolve contradictions on the page. The agent can use web search to compare the sources behind each claim, then update the same page with the best answer or highlight the contradiction for your review.
prompt
resolve contradictions. Search the web, update this page with the best answer, or highlight unresolved conflicts for my review.
The result
Section titled “The result”After a Mana run, the useful result should live in the repo:
- source material is captured or referenced;
- a page is created or updated;
- related pages are linked;
- important claims are grounded;
- contradictions are called out instead of hidden;
- the whole change is committed to git.
That is how Mana keeps a knowledge base organized: by making every edit improve the structure, not only the text.