Menu
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.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
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.
| name | quick-note |
| description | 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. |
Captures a single insight to 0-inbox/ mid-conversation. No preview, no approval, no multi-agent pipeline. One subagent, one note, one confirmation line.
/learning-loop:quick-note: infer the insight from conversation context/learning-loop:quick-note "insight as title": user provides the title/learning-loop:quick-note "title" "body": user provides everythingIf no args: Read the recent conversation. Identify the most capture-worthy insight: a decision made, pattern discovered, or connection drawn. If nothing stands out, tell the user and stop.
If title only: Use it as the insight title. Derive body from conversation context.
If title + body: Use both as provided.
Spawn a single note-writer subagent (subagent_type: learning-loop:note-writer) with this prompt (resolve ${CLAUDE_PLUGIN_ROOT} to a literal path before dispatch — see agents-shared/vault-io.md → Placeholders):
Write a quick inbox note for the Obsidian vault.
Insight: <title>
Context: <body or conversation context summary, 2-3 sentences max>
Source project: <current project name or "conversation">
Date: <today YYYY-MM-DD>
Destination: 0-inbox/
Before writing, run this command to find related vault notes:
node ${CLAUDE_PLUGIN_ROOT}/scripts/vault-search.mjs search "<key terms from insight>" --rerank
Use the top 1-3 relevant results as wiki-links in the note.
Write the note to its destination using the Write tool, then report the exact written path.
Return the filename, title, and the written path when done.
This is a capture, not a promotion: leave destination_locked unset so a genuinely gate-worthy note can still be promoted out of 0-inbox/ by the gate. The note may therefore land in 1-fleeting/ or 3-permanent/ rather than 0-inbox/ — so do NOT assume the folder; read the path the agent reports.
The note-writer is a subagent. Its Write call bypassed PostToolUse, so backlinks and edge inference didn't run. Replay them on the exact path the agent reported (the gate may have routed the note to a folder other than 0-inbox/; a reconstructed 0-inbox/<filename> path would point at a nonexistent file and the replay would silently fail):
printf '%s\n' "$NOTE_PATH" \
| node "${CLAUDE_PLUGIN_ROOT}/scripts/sweep-hook-replay.mjs" --stdin
Where $NOTE_PATH is the absolute path built from the Written: line in the agent's response — never reconstructed from the requested destination. See skills-shared/hook-replay.md for context.
Show one line, using the actual folder the agent wrote to (read it from the agent's reported path):
Captured: "Note Title" → <actual-folder>/filename.md
Nothing else. No summary. No follow-up suggestions. Back to what we were doing.
/deepen exists for a reason.