gaps

name	gaps
description	Use when the user wants to find concepts they should write notes for but haven't, e.g. "what notes am I missing on reinforcement learning", "find gaps in my philosophy notes", "what should I write next". Surfaces terms/concepts referenced across their notes and readings that have no note yet, ranked as a next-write queue. Never writes the notes for them.

Gaps (next-write queue)

Find the concepts the user keeps using but hasn't written up, the holes in their Zettelkasten. The output is a prioritized list of notes they should write next. This skill never writes those notes; the writing is the learning.

Core rule: never author note content. This skill only reads and reports. It does not create note files (not even stubs, concept notes are theirs to write; that's the whole point).

Read the vault profile (.vault-mind/profile.md) first if you haven't this session (paths + frontmatter schema; run /vault-mind:init if missing). "Notes dir", "resources dirs" below refer to the profile's values.

Procedure

1. Scope

• Topic (gaps in my RL notes): restrict to that cluster (by tag/name).
• Whole vault (what should I write next): scan the notes dir broadly.

2. Find the gaps

Three signals, strongest first:

1. Dead wikilinks: [[Concept]] references pointing to notes that don't exist. These are gaps they already flagged themselves by linking to them.

   # from the vault root; substitute the profile's dir names. links FROM note bodies;
   # strip embed/asset and math-matrix noise, trim whitespace:
   rg -oN '\[\[[^]|#]+' <notes_dir> | sed 's/.*\[\[//; s/^ *//; s/ *$//' \
     | rg -vi '\.(png|jpe?g|gif|webp|avif|svg|mp4|pdf|excalidraw)$' \
     | rg -v '^[0-9 ,.]+$' | sort -u > /tmp/linked.txt
   # "exists" must include BOTH notes and resources dirs (source links point into resources).
   # Omit <resources_dirs> if the profile lists none (passing a missing path makes fd error):
   fd -e md . <notes_dir> <resources_dirs> | sed 's#.*/##;s#\.md$##' | sort -u > /tmp/existing.txt
   comm -23 /tmp/linked.txt /tmp/existing.txt    # linked but no note
   

   Filter the result before reporting:
   • Case/whitespace mismatches to an existing note are NOT gaps; they're broken links (e.g. koopman operator → existing Koopman Operator). Route those to /vault-mind:lint, don't list them to write. Detect by case-insensitive re-check against /tmp/existing.txt (rg -ix).
   • Account for an aliases key (if the profile defines one), a "missing" target may exist under an alias. Verify before reporting.
   • Scanning the notes dir only (not resources) keeps author/person links, which live in reading author frontmatter, out of the results. Authors aren't concept gaps.
2. Recurring un-noted terms: concepts named across multiple notes/readings (especially technical terms or names) that have no note and aren't even linked yet.
3. Cluster holes: a topic they clearly engage with where a foundational/connecting concept is conspicuously absent (e.g. a method whose prerequisites have notes but the method doesn't).

3. Present the queue

Rank by how central/recurring the concept is and how much it would knit existing notes together. For each gap:

• Concept name: where it's referenced (which notes/readings, how many times).
• Why it's worth writing, what it would connect or unblock.
• For dead links: note that they already linked to it from [[X]].

Be honest if the vault is well-covered on the topic.

4. Hand off

Suggest they write the top one or two themselves. If they want to learn an unfamiliar gap concept before writing it, point them to an external from-scratch tutor skill (if installed). After they write a note, /vault-mind:connect can wire it in. Create no note files.