Menú
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.
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
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 | deepen |
| description | 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. |
Launches the note-deepener agent to strengthen a single note. The agent assesses maturity, researches gaps scaled to need, rewrites in persona voice, verifies sources, and promotes when ready. Shallow notes get heavy research; deep notes get a light touch.
/deepen <note-name>: target a specific note/deepen: no argument; auto-picks the note most in need of work — the shallowest 0-inbox/ note OR a gate-demoted 1-fleeting/ note (verification markers / source: unverified) that has no other resurfacing path/inbox flags a note as needing deepening (including the fleeting NEEDS-DEEPEN recommendations)This skill emits provenance events for pipeline observability. Run each Bash command silently.
At session start:
node "${CLAUDE_PLUGIN_ROOT}/scripts/provenance-emit.js" '{"agent":"deepen","skill":"deepen","action":"session-start","target":"NOTE_FILENAME"}'
At session end:
node "${CLAUDE_PLUGIN_ROOT}/scripts/provenance-emit.js" '{"agent":"deepen","skill":"deepen","action":"session-end","target":"NOTE_FILENAME","promoted":true|false}'
Per-note tracking is handled automatically by the PostToolUse hook.
No argument (/deepen):
Run auto-pick immediately (the agent picks the note most in need of work — the shallowest inbox note OR a gate-demoted marker-bearing 1-fleeting/ note — no prompting needed). After presenting results, mention the targeted form in one line:
Deepened [note]. To target a specific note:
/deepen "note name".
Argument provided: Proceed immediately.
Launch the note-deepener agent with:
Glob if only a name was given){{VAULT}}/The agent definition is at ${CLAUDE_PLUGIN_ROOT}/agents/note-deepener.md (resolve to a literal path before dispatch — see agents-shared/vault-io.md → Placeholders).
If no note name was provided, pass no note_path: the agent will auto-pick from 0-inbox/ (shallowest) and 1-fleeting/ (gate-demoted marker-bearing notes), preferring an older stuck fleeting note over a thin inbox seed.
The note-deepener is a subagent. Its Write/Edit tool calls bypass PostToolUse, so the deepened note (and any split note in 0-inbox/) misses the hooks/post-tool.js dispatcher (autolink + edge-infer modules).
Run the unlinked-body sweep from ${CLAUDE_PLUGIN_ROOT}/skills-shared/hook-replay.md (read it and execute; seed the candidate list with the destination path — and any split-note paths — from the agent's report, then it backfills via the unlinked-body walk and replays the hook chain on each). Idempotent: safe on already-hooked notes.
Report failures in Step 2 if any.
The agent returns a structured report with before/after comparison, maturity transition, and destination. Present it to the user.
If the agent flagged uncaptured sources, suggest /literature for each.
If the note contains verification markers, prioritize resolving them. The canonical marker vocabulary — which markers exist, which block promotion, and how each resolves — lives in agents-shared/capture-rules.md → Verification Markers. Read that section rather than relying on a local list; restated copies drift. For [unverified], the inspection command is node ${CLAUDE_PLUGIN_ROOT}/scripts/source-resolver.mjs verify-note <path>. Remove each marker only once its resolution rule is satisfied.
note-deepener agent and its agents-shared/.0-inbox/.