Skip to content

The daily workflow

Once a wiki is set up, the loop settles into a rhythm. Drop sources continuously, ingest them daily, lint weekly, archive monthly, and run the read and update commands any time.

CADENCE OF THE DAILY LOOP CONTINUOUS throughout your day drop files into entry points no command needed; files sit until /ingest runs EVERY FEW DAYS check, then compile THEN /status what changed? /ingest distill new sources pages written, manifest updated WEEKLY maintenance THEN /lint find issues /cross-linker repair wikilinks orphans, broken links, stale pages MONTHLY OR RARE snapshots and resets /archive snapshot /rebuild reprocess all sources rebuild only after schema changes AD-HOC, ANY TIME use, refine, teach /query ask the wiki /update refine a page /research web + distill /capture save this chat /ingest-url one URL /ingest-claude conversation exports /project create/archive /feedback teach the agent LEGEND primary loop supporting rare / conditional
  • Continuous: drop files into entry points whenever something is worth keeping. No command needed; the files sit until you ingest
  • Daily or every few days: /status to see what changed, then /ingest to compile the new sources into pages. This is the primary loop
  • Weekly: /lint to surface issues, then /cross-linker to repair link problems
  • Monthly or before a big change: /archive to snapshot the knowledge. Use /rebuild only when you have changed the schema and want to reprocess everything from scratch
  • Any time: /query to ask a question, /update to refine a page, /research to pull sources from the web, /capture to save the durable parts of the current conversation, /capture --quick to stage findings in under 60 seconds, /feedback to teach a new rule
  • Session end (optional): install .claude/hooks/wiki-stop-capture.sh (see wiki-setup/SKILL.md, “Optional: session-end capture hook”) to have Claude Code nudge you with /capture --quick at the end of a session that had edits or shell activity