vault-review
Skill2 starsChanged 7 months agoMITView on GitHub
Walk the human-review queue — pages in Sources/, Concepts/, Entities/ with `explored: false`. Use when the user says "review the queue", "work through pending pages", "check what the agent created", or asks for a review session. Applies LLM judgment per page — not just a bash menu — to approve, flag contested claims, propose edits, or spot duplicates. Do NOT use for ingest (use vault-ingest) or querying (use vault-query).
1 other project uses a version of this.
Adopted by
Most recently taken up by
- sjkeller/second-brain-templatetheir own version, first seen Oct 9, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- sjkeller/second-brain-template★ 2same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository11 other files it gives its agents.
- AGENTS.mdAGENTS.md
- agent-browserSkill
- defuddleSkill
- json-canvasSkill
- obsidian-basesSkill
- obsidian-cliSkill
- obsidian-markdownSkill
- vault-ingestSkill
- vault-initSkill
- vault-maintainSkill
- vault-querySkill
The file
We split this skill into the rules it gives an agent
9 rules. About 0 are common, 7 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 18
After a vault-ingest session the user wants to validate
- Distinctive: line 64
Mark duplicate: if the page is a duplicate of an existing one:
- Very specific: line 66
Add the duplicate's title to the canonical page's aliases:
- Distinctive: line 67
Update all inbound links to point to the canonical page
- Distinctive: line 68
Delete the duplicate (this is the one case where deletion is OK — duplicates have no history worth preserving)
- Distinctive: line 71
Skip: leave as-is; note that it stayed in the queue.
- Distinctive: line 75
After the session, update hot.md's "Pending Review" section to reflect the new count.
- Distinctive: line 95
Prefer the LLM-driven workflow for substantive review.
- Distinctive: line 99
Don't bulk-approve: If the user says "just approve everything," refuse — the point of review is judgment.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Vault Review Skill
Walk the human-review queue with actual judgment. Every AI-created page starts with explored: false; a human flips it to true after reviewing. This skill helps the human reviewer by doing the heavy reading and pointing out what to notice.
When to Use
Trigger on any of:
- "Review the queue"
- "Walk through the pending pages"
- "What needs review?"
- "Check what you created"
- After a
vault-ingestsession the user wants to validate
When NOT to Use
- The user wants to read a specific page — just read it
- The user wants bulk approve without reading — that defeats the point; refuse and explain
- The queue is empty — report
bin/review.sh --countand stop
Workflow
Step 1 — List the queue
bin/review.sh --listThis returns all files in Sources/, Concepts/, Entities/ with explored: false. If empty, report and stop.
Step 2 — Walk each page with judgment
For each page in the queue, don't just display it — read it and report:
- Show the page (frontmatter + first 40–50 lines of body)
- Spot-check:
- Are Concepts/Entities properly cross-linked back to the Sources/ page?
- For Concepts/: are Counter-arguments and Data Gaps sections present and non-trivial?
- Does
confidencematch the evidence (multiple sources = high, single source = low)? - Is this a duplicate of an existing page (check aliases, similar-titled pages)?
- Any claims that look inferred but aren't marked
> [!info] Inferred? - Any
status: high confidencepaired with thin evidence?
- Propose an action: approve / approve-with-edits / contest / mark-duplicate / skip
- Wait for user to confirm or adjust
Step 3 — Apply the action
Approve (frontmatter flips): use Edit to change explored: false → explored: true. Optional: bump status (seed → developing if 2+ sources) or confidence (medium → high if well-supported).
Contest (keep explored: false, add a note): append:
> [!question] Contested (review YYYY-MM-DD)
> <reviewer's note>Cross-link to a new or existing Questions/ page if the contest is substantive.
Mark duplicate: if the page is a duplicate of an existing one:
- Merge useful content into the canonical page
- Add the duplicate's title to the canonical page's
aliases: - Update all inbound links to point to the canonical page
- Delete the duplicate (this is the one case where deletion is OK — duplicates have no history worth preserving)
- Report clearly what was merged
Skip: leave as-is; note that it stayed in the queue.
Step 4 — Update hot.md
After the session, update hot.md's "Pending Review" section to reflect the new count. If the user approved or contested a notable page, note it in "Recent Changes."
Step 5 — Report session summary
Reviewed: N pages
Approved: X
Contested: Y
Duplicates merged: Z
Skipped (remaining): W
Queue size: before → afterNon-Interactive Shortcuts
bin/review.sh --count— queue size onlybin/review.sh --list— list of filesbin/review.sh— bash-driven interactive mode (menu-based, no LLM judgment)
Prefer the LLM-driven workflow for substantive review. The bash script is for when the user wants fast menu-driven approvals.
Anti-Patterns
- Don't bulk-approve. If the user says "just approve everything," refuse — the point of review is judgment.
- Don't flip
exploredtotruewithout reading. Frontmatter drift is a silent failure mode. - Don't create new pages during review. This is a verification loop, not an authoring loop. If review reveals a missing Concept, flag it for the next ingest or query — don't sneak it in.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

