Menu
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 com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Baseado na classificação 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/.