agentleFS
erikunhaerikunha/vault-of-loci/.agents/skills/review-queue/SKILL.md

review-queue

Skill0 starsChanged 5 days agoMITView on GitHub

List what is next to review from the note frontmatter and the Obsidian Bases (weakest concepts, concepts not tested in 14 days, unresolved links), within a daily review cap and with a catch-up mode for a backlog, optionally for one track. Use when the learner asks what to review next or which concept is weakest or overdue. Reads by default and sets properties only on request.

  • Reads credentials

8 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Files
SKILL.md

Tutor mode: this skill lists; the learner decides and attempts. It never answers questions, writes the study layer or sets a score the learner did not report.

With the Obsidian CLI

Needs the app running and Obsidian 1.12.7 or later. Check with obsidian help; if the command fails, use the file fallback.

obsidian base:query path="90_System/Bases/Weakest.base" format=json
obsidian base:query path="90_System/Bases/Interleave.base" format=json
obsidian base:query path="90_System/Bases/Progress.base" view="Review next" format=json
obsidian unresolved

INDEX.md links every note, so obsidian orphans lists nothing useful here. Set a property only when the learner reports the value:

obsidian property:set name="last_tested" value="2026-11-07" file="Retrieval_Practice"

Parameter names can change between releases: obsidian help base:query is the reference.

From files

Read the frontmatter of each note in 20_Concepts (never .env or plugin data.json). Sort by score ascending, then last_tested ascending; empty values come first. The "Review next" view of Progress.base ranks by (5 - confidence) x (days since last test + 1), so a weak concept not seen for a while comes first; an untested concept counts as 30 days. For one track, keep the concepts whose tracks property links to it. Find unresolved links with check_links.py.

Calibration and staleness

python3 -I 90_System/Scripts/calibration_report.py lists concepts where the learner's confidence and score are 30 points or more apart, concepts not tested for 30 days and concepts never tested. Put the overconfident ones first: they feel known and are not.

Daily cap and backlog

A day's flashcard work is 10 to 20 reviews and about 2 new cards (defaults from 90_System/Specs/Track_kinds.md, not research results). Ask the learner how many cards are due.

  • Due at or under the cap: normal review, new cards allowed.
  • Due over the cap: catch-up mode. Review the oldest and weakest first, stop at the cap, add no new cards until the queue is under it. Missed days carry no penalty and no streak resets.

Output

At most 5 concepts, each with score, last_tested and confidence. Mix categories: take the weakest first, then alternate categories (interleaving). Say whether today is normal or catch-up mode. End with one suggestion: no-ai-checkpoint for the list, or mistake-review when 50_Review/Mistakes.md has open entries.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.