/study-unstudied — What hasn't been studied yet

Live query over sources/ + concepts/ + study-sessions/. No persisted queue — regenerated on every invocation so state is always accurate.

Flow

1. Enumerate sources. List every sources/YYYY-MM-DD-*.md file (exclude README.md, INDEX.md).
2. Enumerate concepts. List every concepts/<slug>/ subfolder with a README.md.
3. For each source — check if study-sessions/<source-filename>.md exists. If not → unstudied source.
4. For each concept — check if concepts/<slug>/study-log.md exists AND has at least one ## Checkpoint — header. If not → unstudied concept.
5. Score + rank unstudied items:
   • Read each unstudied source's matching brainstorm from brainstorming/ (filename pattern *<source-slug>*)
   • Count convergence-tag membership (from brainstorm frontmatter convergence: list) — higher count = more leverage
   • Sources in clusters of 3+ brainstorms (per scripts/kb-convergence-scan.py threshold) rank first
   • Concepts rank by number of referenced KB signals in their README.md
6. Render:

## Unstudied

### Sources (<N>)

**High-leverage cluster: <cluster-tag>** (<N> brainstorms converge here)
- [[sources/YYYY-MM-DD-<slug>]] — <short description>
- ...

**Other unstudied:**
- [[sources/...]] — ...

### Concepts (<N>)
- [[concepts/<slug>]] — <description>
- ...

Keep the list scannable. Full /study-status output belongs in that skill — here, one line per item.

Voice mode stub

--mode voice: same query, renderer narrates "you have 23 unstudied sources, most clustered around <top-cluster> — top three are <X>, <Y>, <Z>. Want to start with one of those?"

What /study-unstudied is NOT

• Not a writeable queue. If you want to pin specific sources for later, add them to todos/ via /add-todo.
• Not a ranking of importance. Convergence-tag count is a proxy for leverage, not intrinsic value — a standalone source can still be worth studying.
• Not a replacement for /kb-audit. That skill flags stale/broken things; this one flags unstudied ones.