Vault Agents

A suite of automated agents that run nightly (3am) to maintain vault health, keep documentation organized, and surface useful information.

Goals

• Maintain documentation integrity across all vaults
• Keep notes organized and cross-linked
• Surface reminders and outstanding action items
• Synthesize information over time

Agent Ideas

Core Maintenance

1. Cleanup Agent Scan all vaults for formatting issues, stale placeholders (e.g. TBD, TODO, entity IDs not yet confirmed), and frontmatter gaps. Flag or auto-fix where safe.

2. Staleness Detector Flag notes that haven’t been updated in X days, or that still contain unresolved placeholders from their creation. Produces a report or updates a “needs attention” section.

3. Tag Auditor Normalize inconsistent tags across vaults. Suggest missing tags based on content. Optionally auto-apply.

Synthesis & Linking

4. Cross-Vault Linker Identify concepts, terms, or topics that appear across multiple vaults. Suggest or insert [[wikilinks]] to connect related notes.

5. Action Item Extractor Scan all notes for loose TODOs, next steps, and outstanding items buried in content. Consolidate them into the todos vault automatically.

6. Daily Digest Generator Produce a short summary note of what changed across all vaults in the last 24 hours. Good for a morning review.

9. Consistency Review Agent ← in planning Read all notes within a vault, check for internal consistency (conflicting facts, outdated references, terminology drift, structural gaps). Auto-apply safe corrections and log them. When confused or when a decision requires user input, write action items to the todos vault. See consistency-agent for full plan.

Reminders & Awareness

7. Reminder Agent Read all vaults for time-sensitive items, deadlines, or flagged follow-ups. Output a daily reminder note or push a notification.

8. Orphan Detector Find notes that exist in a vault but are not linked from any other note. Surface them so they can be integrated or discarded.

Shortlist — Pick 1–3 to Plan First

Select below. Each chosen agent will get its own planning note with: purpose, trigger, inputs, outputs, logic, and open questions.