learning-loop

Deep research with the local librarian doing the token-heavy middle. Scope on Claude, Search+Fetch+Extract on the local Ollama model, adversarial Verify + Synthesize back on Claude. Usage: /learning-loop:research "<question>". Falls back to Claude-native WebSearch when the librarian is unavailable or sub-tier.

End-of-session consolidation. Usage: /learning-loop:reflect (no args). Reviews conversation, extracts learnings, routes to auto-memory or vault, cross-links projects, promotes inbox notes. Run after substantial work sessions.

measure whether a /dream pass helps, hurts, or ties against no consolidation. usage: /learning-loop:dream-eval --mode={single|control|repeated} [--passes=N] [--mine]. single scores pre/post on the live corpus; control forks two clones and dreams one; repeated dreams a clone N times and plots the drift + content-survival curve.

Consolidate auto-memory between sessions. Usage: /learning-loop:dream (no args). Four-phase cycle: Orient, Gather Signal, Consolidate, Prune Index. Seven operators: MERGE, RESOLVE, ABSTRACT, COMPRESS, PRUNE, LINK, DATE NORMALIZE. Rebuilds MEMORY.md.

Cross-store correction for a retracted or updated belief. Usage: /learning-loop:rewrite "old pattern" "new pattern" [reason]. Searches vault, auto-memory, and episodic history; presents an impact map for triage; executes user-approved changes; records a supersession so future episodic searches get annotated.

Diagnose your learning-loop installation. Runs health checks, presents issues, offers per-fix remediation, re-runs each check after the fix to confirm. Safe to run anytime; only makes changes you approve. Pass --redact to scan plugin data files for leaked credentials instead.

Show all learning-loop commands with usage and modifiers. Usage: /learning-loop:help (no args).

Pull external context into the second brain. Handles any format Claude can read: PDFs, images, code, conversations, docs, or raw text. Usage: /learning-loop:ingest linear ["project"], /learning-loop:ingest repo [path] [--deep] (--deep forces the parallel deep-mapper tier), /learning-loop:ingest context, /learning-loop:ingest bundle <path> (restore a harvest bundle), /learning-loop:ingest (prompts for source); add --refine to any source mode to run upstream refinement after ingest.

Explore a topic interactively with web research and vault context. Usage: /learning-loop:discovery "topic" [--style guided|branch|checkpoint] [--tone academic|conversational|persona] [--surf]. --surf skips note creation.

Set up or repair learning-loop federation: identity creation, token redemption, Tailscale, visibility rules, and the first sync test. Run when /init asked you to defer federation, when sync is broken, or when rotating identity. Safe to re-run.

Challenge vault knowledge through coverage, confidence, and counterpoint analysis. Usage: /learning-loop:gaps "topic" | /learning-loop:gaps (auto-pick) | /learning-loop:gaps --sweep; any mode takes [--depth shallow|medium|deep] [--dry-run]. Scales depth to note maturity (--depth overrides); --dry-run shows what would be analysed without researching. Creates counterpoint notes, offers rewrites.

Collect opt-in-marked, IP-scrubbed insights from THIS instance to carry back to another learning-loop instance you own. Usage: /learning-loop:harvest [--all] [--out <dir>]. Keys ONLY on frontmatter `portable: true` (never `visibility`). Mechanical deny-list blocks IP; you review what survives. Emits a lift-bundle to ingest on the home instance.

Vault health dashboard. Usage: /learning-loop:health [--deep] [--auto] [--provenance] [--librarian]. Light mode (default) shows counts + file lists. --deep uses note-scorer for full analysis. --auto fixes safe issues without asking. --provenance shows pipeline observability; --librarian reviews queued librarian suggestions.

Batch triage inbox notes and sweep fleeting for archival. Usage: /learning-loop:inbox. Classifies intention status (intentioned/resolved/limbo), clusters by topic, auto-promotes mature notes, surfaces top-5 limbo notes for close-or-plan decision, sweeps 1-fleeting/ for promoted/stale notes to archive, recommends merge/deepen/delete for the rest (asks before destructive actions).

First-time setup or upgrade for the learning-loop plugin. Configures vault path, persona voice, CLAUDE.md integration, and verifies the installation. Hands federation off to /learning-loop:federation. Safe to re-run: detects existing state and skips completed steps.

Fast answer to a question with vault context (the spoken answer is ungated; only captured notes are verified before writing). Usage: /learning-loop:quick "question" or /learning-loop:quick (infers from context). One-shot: web research + vault awareness + auto-capture if novel.

Build a portable starter slice from this learning-loop instance for a fresh instance to boot from. Usage: /learning-loop:seed [--for-job] [--types feedback,reference] [--out <dir>]. Default --for-job carries working-style (feedback memories + _system + portable prefs), no vault notes, no projects. Use when ramping up at a new job, a new project vault, or a second machine.

Cleanly remove learning-loop and its footprint. Usage: /learning-loop:uninstall. Walks the three steps — marketplace removal, episodic-memory MCP removal, and purging captured indexes — with confirmation before each destructive action. Does not auto-delete.

Assess note quality and verify claims against cited sources. Usage: /learning-loop:verify "note-name" | inbox | fleeting | permanent | "topic" (defaults to inbox). Scores quality, checks source integrity, detects duplicates: produces fix plan.

Strengthen a single vault note with research. Usage: /learning-loop:deepen "note name" or /learning-loop:deepen (picks shallowest inbox note). Assesses maturity, researches gaps, rewrites in vault voice, promotes when ready.

Capture an external source as a literature note. Usage: /learning-loop:literature <URL> or /learning-loop:literature "paper title". Fetches content, extracts core ideas in vault voice, writes to 2-literature/ with vault links.

Quick mid-conversation capture to inbox. Usage: /learning-loop:quick-note (infers from context), /learning-loop:quick-note "insight title", or /learning-loop:quick-note "title" "body". Writes to 0-inbox/ without breaking flow.

Generate an Excalidraw diagram for the vault. Usage: /learning-loop:diagram "concept or mechanism" or /learning-loop:diagram (infers from context). Produces .excalidraw.md in vault Excalidraw/ folder.

See what you already know about a topic (no web research). Usage: /learning-loop:refresh "topic". Searches vault notes, episodic memory, and literature. Optionally tests discrimination of confusable note pairs. Good before /learning-loop:discovery or when returning to a topic.