Memory curator

Memory Curator

Turn a session's raw bd remember notes into deduplicated, consolidated, well-structured memories — and prune the pile — using bd over text already in context. No runtime, no embeddings.

Announce at start: "I'm using the memory-curator skill to consolidate and structure the memory store."

When to Use

• Session-close — when the session produced ~3+ new bd remember calls. Offered, never automatic.
• On-demand — a full-store sweep: Skill(memory-curator).

When NOT to Use

• Sessions with 0–2 new memories — not worth a pass.
• Mid-task — run at a clean stopping point, not while work is in flight.

Memory taxonomy

Two classes; procedural memory (how-to / workflow) lives in the skills, never the memory store.

• semantic — durable facts that stay true. Subtypes: design, lesson, pattern, decision, root-cause, research, correction. The durable core of the store.
• episodic — time-bound records of what happened. Subtypes: done, continuation, cleanup, review. Consolidate and retire as they age; a cluster often distills into one semantic fact.

Map a non-canonical prefix to the nearest canonical subtype — e.g. stress-test/plan-stress-test→design, bug→root-cause, sdd→lesson, upstream→research, docs→pattern. If none fits, ask — don't invent. If an extracted "memory" is really procedural, flag it for a skill — don't store it.

Memory header

Every memory keeps its existing key and carries one greppable header line:

@type=semantic:lesson @created=2026-06-28 @salience=4 @refs=<bead-id>,<memory-key> @tags=memory,curation
<self-contained fact body>

• @type — <class>:<subtype> from the taxonomy. @created — ISO date. @salience — 1–5, best-effort. @refs — related bead IDs / memory keys. @tags — lexical filter.

One line. The class makes the prune signal greppable (bd memories | grep '@type=episodic:'); @salience/@tags filter recall.

The sweep

One pass. Input: the session (in context) + bd memories --json. Output: a reviewed list of bd remember / bd forget commands. Propose least-destructive changes first (enrich + exact-duplicate dedup); cross-cluster consolidation and pruning come after, and only where clearly safe.

1. Gather — only after an explicit non-mutating offer is accepted, use bd memories --json for the full store; bd dolt status records the pre-sweep state for rollback.
2. Extract — pull salient, self-contained, date-grounded facts; classify each by the taxonomy and normalize its @type to class:subtype (correcting any malformed @type it encounters). Store a fact ONLY if it carries checkable evidence (cited file:line, passing test, command output, closed bead) — the same bar as Agent-Filed Bead Discipline in verification-before-completion. No evidence → do not store. Procedural how-to → flag for a skill, don't store. Never persist secrets, credentials, tokens, keys, or PII — the session hook injects curated memories into every future session (the full store via bd prime) and Dolt history outlives bd forget.
3. Reconcile — ADD new facts; UPDATE a same-topic memory in place with bd remember --key <existing>, merging so the result keeps the MOST information (never silently shrink); skip what's already present.
4. Consolidate — collapse a themed cluster of episodic memories into one timeless semantic fact with @refs to its sources, then retire the cluster. The only step that shrinks the pile. Extract a record's durable content into a semantic memory BEFORE retiring it — never drop an episodic record that still holds an un-consolidated fact.
5. Forget — soft-tombstone a superseded memory ([superseded YYYY-MM-DD by <key>]) rather than delete — Dolt keeps history either way, and a tombstone is reversible if the supersede was wrong. Episodic records are the prune-first candidates, but never retire the most-recent continuation / active handoff. Reserve hard bd forget for exact duplicates or true noise, with a cited reason.

Iron rule: propose, then apply

This mutates the store injected into every future session (curated by the session hook, in full via bd prime) — a bad run corrupts the context layer invisibly. So:

• Emit the full planned command list — every ADD / UPDATE / CONSOLIDATE / FORGET with a one-line reason — and get the user's approval before running ANY of it. The on-demand sweep is dry-run-first, always.
• Surface the pre-sweep Dolt state (step 1) as the rollback path.
• No hard bd forget without an exact-duplicate match or a cited supersede reason.

Red Flags

Thought	Reality
"I'll just apply the merges"	Never mutate silently. Propose the list; the user approves first.
"This memory is probably fine to store"	No cited evidence → it doesn't meet the bar. Drop or low-salience.
"I'll keep the shorter version on UPDATE"	UPDATE keeps the MOST information. Never silently shrink.
"Consolidate hard to shrink fast"	Lose no distinct fact; extract durable content to semantic first, then retire.
"Episodic, so safe to drop"	Never retire the latest continuation/handoff; soft-tombstone, never hard-delete.
"There might be a token in here, but it's internal"	Never persist secrets/PII. Redact or skip.

Beads Integration

bd create "Memory curation: <session/sweep>" -t chore
# after the user approved + you applied:
bd close <id> --reason "Curated: <N added, M updated, K consolidated, J forgotten>; pre-sweep Dolt <ref>"

Run this as the session/ledger-owning agent; a dispatched single-task subagent does not.

Integration

Invoked at: session-close (offered when a session produced ~3+ new memories — see finishing-a-development-branch Step 7) and on-demand by the user. Pairs with: verification-before-completion (supplies the evidence bar) and getting-up-to-speed (its session-start bd forget is lightweight cleanup; this skill owns curation).

Memories arrive header-less from other skills; the curator assigns @type on contact. Do not add @type emission to other skills — header-less-until-curated is the intended state.

The Capture Gate rejects approval/completion episodes, procedural recipes, raw failure output, current branch/HEAD/next-task state, and artifact pointers before storage. Curation does not weaken that bar. Preserve only the latest relevant continuation when it has an explicit @expires=YYYY-MM-DD; expired continuations are retirement candidates after user approval.