Maintenance Checklist¶

Literal checklists by cadence. Print these, stick them somewhere visible, tick them off. The longer-form reasoning for each item is in 12-maintenance.md.

• • •

Daily (2 minutes, with coffee)¶

☐ Open this morning’s _lint/lint-YYYY-MM-DD.md report.
☐ If there are error-severity findings: fix them before the workday starts.
☐ Skim .memex/log.jsonl for any lint_daily_*_error events; investigate if present.

Nothing else is required daily.

• • •

Weekly (60–90 minutes, same day and time every week)¶

Inbox sweep — 15–20 min¶

☐ Open inbox/.
☐ For every item: promote to raw/ + delete inbox item, merge into an existing wiki note + delete inbox item, or delete outright.
☐ inbox/ is empty at end of this step.

Raw triage — 10–15 min¶

☐ For every new raw/ file this week: add processed: true/false and, if false, atomize: true/false or equivalent annotation.
☐ No raw file is left in ambiguous state.

Compile queue — 15–20 min¶

☐ For every raw file marked compile-now this week: memex compile <path> (dry-run).
☐ Review each dry-run against the source.
☐ memex compile <path> --apply for those whose plans are correct.
☐ Open each newly compiled wiki note; read it; edit weak prose.

Link pass — 10–15 min¶

☐ Filter wiki for status: seed files created this week.
☐ Each has ≥ 2 outbound [[links]].
☐ Create stub notes for any link targets that don’t yet exist.

Lint report review — 5–10 min¶

☐ Open every _lint/lint-*.md report from the week.
☐ Error findings: zero remaining (fix now if any).
☐ Warning findings (orphans, duplicates): each investigated — link, merge, or note as future work.
☐ Info findings (stubs, missing latticework): two most interesting stubs scheduled for development.

Close — 5 min¶

☐ memex lint exits clean.
☐ git add . && git commit -m "weekly review YYYY-MM-DD" in the vault repo (if versioned).

Gate. If lint is not clean and you have no more time, note the blocker somewhere visible; do not declare the review done.

• • •

Monthly (2–3 hours, first working day of the month)¶

Duplication sweep — 30 min¶

☐ memex lint — read all orphan warnings.
☐ For each orphan: decide link, merge, or archive.
☐ For merges: pick the stronger slug, copy content, find-and-replace inbound links, delete loser, run memex lint to confirm clean.

Promotion pass — 30 min¶

☐ Wiki filter: status: seed older than 2 months.
☐ Each: promote to draft, merge, or archive.
☐ Wiki filter: status: draft older than 3 months and used in at least one output.
☐ Each: promote to evergreen (after re-validating claims) or leave as draft with reason.

Ontology refresh — 30–45 min¶

☐ Walk every note in people/, companies/, philosophies/, eras/.
☐ Touch updated: field honestly.
☐ If entity’s status changed, update the body.
☐ Archive entities that are no longer active (status: archived + move to archive/<subfolder>/).
☐ New entities that emerged this month: create with full context now.

Coverage check — 15 min¶

☐ memex chart latticework-coverage --apply.
☐ memex chart tag-frequency --top 20 --apply.
☐ memex chart type-distribution --apply.

☐ Open each chart. Identify one or two targeted actions based on gaps; add them as inbox items.

Project close-outs — 15–30 min¶

☐ For every project with target_end_date past: promote generalisable notes to wiki; archive project folder.

Backup — 5 min¶

☐ Off-device backup completed and verified (rsync / remote git push / Obsidian Sync up to date).

• • •

Quarterly (4–6 hours, first weekend of the quarter)¶

The essay pass¶

☐ Choose one major topic from your long-running themes.

☐ memex export essay "<topic framed as a position>" --apply.

☐ Read the first draft.
☐ Edit aggressively until it’s publication-quality.
☐ During editing, note every gap: missing concepts, weak claims, missing entities, missing links.
☐ After editing: create every stub the essay wanted to link to; fix every weak note; write every new atomic note the essay surfaced; promote every note that proved itself.

• • •

Yearly (1 day, end of calendar year)¶

Ontology audit — 3 hours¶

☐ Every people/ note read.
☐ Every companies/ note read.
☐ Every philosophies/ note read.
☐ Every eras/ note read.
☐ Each: updated: touched honestly; archive if no longer active; new context added if relevant.

Tag taxonomy review — 1 hour¶

☐ memex chart tag-frequency --top 50.

☐ Tags used < 3 times: merge into a more general tag or kill.
☐ Near-duplicate tags: decide on canonical form; rename everywhere.
☐ Flat tags that should be namespaced: add prefix; rename.

Schema review — 1 hour¶

☐ Read memex/schemas.py end to end.
☐ Note fields you wish existed.
☐ Note fields unused in the past year.
☐ Write next year’s schema wish list in _essays/schema-wishlist-YYYY.md.

Archive cleanup — 1 hour¶

☐ Walk archive/.
☐ Confirm no active note links into archive/ (linter enforces; double-check).
☐ Delete snapshots older than a year whose migrations are still valid.

Infrastructure check — 1 hour¶

☐ Engine on latest stable version.
☐ All Obsidian plugins current.
☐ Python up to date.
☐ API key valid; no leaked copies in any repo.
☐ Backups running and restorable (do a test restore of one file).

Year-in-review — 1 hour¶

☐ _essays/year-in-review-YYYY.md:

☐ What the vault accomplished this year.
☐ Where it failed you.
☐ What you’ll tend differently next year.

• • •

Red-flag checklist (run if anything feels off)¶

Quick diagnostics when you suspect decay:

☐ Is inbox/ empty? (If non-empty for > 1 week: decay.)
☐ Does memex lint exit clean? (If no: decay.)
☐ Did the daily cron run this morning? (grep lint_daily_complete .memex/log.jsonl | tail -1.)
☐ Has any new atomic note this week been used in an output? (If no for several weeks: output habit has lapsed.)
☐ Last weekly review completed on schedule? (If two missed in a row: the system is drifting.)
☐ Last monthly refactor done within the last five weeks? (If not: duplication is accumulating.)
☐ Backups ran in the last seven days? (If not: fix before anything else.)

Any one of these is a yellow flag. Two is red. Three means stop creating, start tending.

← Quickstart · Docs index · Glossary →