Workflow Vault Weekly

Skill: /workflow-vault-weekly

Runs the five vault-health skills in dependency order so the later stages have current data from the earlier ones. Use weekly.

When to use

• Friday-afternoon vault review.
• After a week of heavy note capture, before reviewing for patterns.
• The user explicitly says "weekly vault review" / "what should I publish this week".

When NOT to use

• Daily; the daily rituals (/morning, /eod, /log) cover the within-day cadence.
• For specific known issues (one orphan note, one stale tag) — call the targeted skill directly.

Steps

Stage 1: Lint — step lint — invokes vault-lint

Scans Concepts/, finds orphans, broken links, missing pages, contradictions. Writes vault-lint-report.md.

Output: report file exists. Checkpoint: surface any blocking issues before continuing.

Stage 2: Backlinks — step backlinks — invokes backlinks

Audits and repairs the backlink graph; flags structurally important orphans.

Output: backlink-audit summary in conversation. Checkpoint: confirm fixes applied before rebuilding the index.

Stage 3: Vault index — step index — invokes vault-index

Rebuilds the vault-index note from current state.

Output: vault-index.md updated. Checkpoint: skip checkpoint by default — auto-continue.

Stage 4: Weekly signals — step signals — invokes weekly-signals

Reads patterns.md + the week's daily notes; surfaces deferral patterns and flagged items.

Output: weekly-signals output (in conversation; can be saved if user asks). Checkpoint: confirm signal review before surfacing publishable threads.

Stage 5: Weekly learnings — step learnings — invokes weekly-learnings

Looks for publishable threads from the week's signals + vault deltas.

Output: list of 1–3 candidate threads.

Resume protocol

Default: run lint → backlinks → index → signals → learnings in order, with checkpoints between each (except index, which auto-continues).

--resume-from <step>: skip every step before <step> and begin from <step>. The named step itself runs.

If <step> is not in {lint, backlinks, index, signals, learnings}, print:

Invalid resume-from step: <step>
Valid steps: lint, backlinks, index, signals, learnings

…and stop.

Required state when resuming

The vault stages mostly read from the live vault, so missing prior outputs are usually fine — but a few steps depend on prior reports.

• --resume-from lint — same as a fresh run.
• --resume-from backlinks — runs against the live vault. If vault-lint-report.md is absent, warn the user that lint would normally have surfaced issues first; ask whether to continue.
• --resume-from index — runs against the live vault. No required prior state.
• --resume-from signals — needs patterns.md and at least the current week's daily notes. If patterns.md is missing, surface the gap and stop.
• --resume-from learnings — works best with the previous step's signals output. If signals weren't run this week, ask the user to paste the signals output or confirm we should derive learnings without them.

Examples:

/workflow-vault-weekly --resume-from signals --no-pause  # cron-style mid-week signal pull
/workflow-vault-weekly --resume-from learnings           # signals already reviewed manually

Other flags

• --skip <step> to skip a step (e.g. --skip learnings to stop after signals).
• --no-pause to run end-to-end (use this for the cron-driven weekly run).

Stage quality gate

After each stage: does the output actually address the stage goal (lint = issues enumerated; backlinks = graph repaired; index = current; signals = patterns surfaced; learnings = candidate threads identified)? If it partially misses, retry the stage once with an explicit note on the gap. On second failure, surface to the user and move on rather than blocking the rest of the pipeline.

Failure handling

If a stage errors, stop and print: which step failed, the underlying skill's error, and the resume command (/workflow-vault-weekly --resume-from <step>).