Record what you did, any choices you made, and anything you discovered. Call before committing or ending a session. The summary becomes a status post. Decisions become tracked records with rationale. Findings become blackboard entries visible to future agents. Scope is auto-inferred from git diff if omitted.

Run periodic maintenance on Twining stores. Preview by default (dry run); preview simulates the full pass pipeline, so its counts match what execute will do on the same state (#39). Removes duplicates, surfaces stale decisions and dangling warnings, prunes orphaned graph entities, rotates old metrics, and backfills missing superseded_by back-links on superseded decisions. The blackboard archive pass is OPT-IN (archive: true): it sweeps every archivable entry regardless of age, keeping decisions, unresolved need/warning/question entries (#40), and the newest archive.retain_recent entries (D4). Pass staleness_review: true to also flag entries whose scope/files/branch are gone. Pass execute: true to apply changes.

Archive a list of stale items by ID — typically the candidate IDs returned by twining_housekeeping with staleness_review: true. Decisions move to status "archived" (excluded from assemble/why; reversible via twining_unarchive). Blackboard entries are DELETED from the live board, with a tombstone (full entry + reason) appended to .twining/archive/ — twining_unarchive does NOT restore them.

Restore archived decisions — the undo for twining_archive_stale. Each decision returns to its PRE-ARCHIVE status (archived_from): a provisional goes back to the ratification queue, a superseded decision stays retired, and only previously-active decisions become authoritative again. Records archived by a pre-2.7 server carry no archived_from marker — those restore to "active" as an ASSUMPTION, reported per-id in assumed_active and via a warning post (if one was provisional, that restore ratified it; re-check with twining_reconsider). Archived decisions are excluded from assemble/why (assemble reports them as archived_excluded_count). Only decisions currently in status "archived" are restored; other IDs are reported back untouched.

Share a finding, warning, need, or status update with other agents. Post a 'status' entry before ending each session. Does NOT accept entry_type 'decision' — use twining_decide instead.

Mark open blackboard items (needs, questions, warnings) as handled. Persists status "resolved" with resolver identity and an optional note; the entry leaves the open triage/assemble lane but stays on the board as searchable history. This is the everyday exit for open items — use twining_dismiss only for noise that should never have been recorded.

Before modifying a file, check what decisions constrain it. Shows rationale and alternatives so you don't contradict prior choices. Results are ranked by relevance and bounded by a token budget; overflow decisions appear as one-liners in more — pass their ids back via ids for full detail.

Your FIRST call every session. Returns a briefing with decisions to respect, warnings to address, and handoff context from previous agents. Call BEFORE reading code or making changes. token_estimate ≈ max_tokens is the signature of budget truncation: the briefing (and decisions_count) was clipped — re-call with a larger max_tokens (e.g. 100000) for complete coverage. decisions_count is the briefing selection, not a scope census; use twining_why total_in_scope for populations.

Overall health check of the Twining state. Shows blackboard entry count, decision counts, graph entity/relation counts, actionable warnings, the server_version and resolved storage backend, and a human-readable summary. provisional_decisions is the canonical ratify-queue count — a direct index count no query can distort (scoped variant: twining_triage counts.open.by_kind.decision). Note: twining_assemble now includes a status summary — use this only when you need the full detailed health check.

Archive old blackboard entries. Moves entries older than a cutoff timestamp to an archive file, preserving decision entries and unresolved need/warning/question entries (#40 — an item counts as resolved when explicitly resolved via twining_resolve or when a later entry references it via relates_to). Optionally posts a summary finding. WARNING: the cutoff defaults to now, so an argument-free call archives everything archivable — pass before or retain unless a full sweep is intended.

Add or update a knowledge graph entity. Uses upsert semantics: if an entity with the same name and type exists, its properties are merged and updated. Returns the entity ID.

Add a relation between two knowledge graph entities. Source and target can be entity IDs or names. Returns an error for ambiguous name matches. Upsert semantics: re-adding the same (source, target, type) merges properties instead of duplicating the edge. Relations are provenance-marked: agent-typed edges get properties.origin "declared", auto-populated edges "derived", absent means legacy/unknown.

Traverse the knowledge graph from an entity, returning neighbors up to a given depth (max 3). Supports filtering by relation type. Useful for understanding how entities connect.

Search the knowledge graph for entities by name or property substring match. Case-insensitive. Returns matching entities with their properties.

Remove orphaned knowledge graph entities that have no relations. Use this to clean up stale or disconnected entities. Optionally filter by entity type to only prune certain kinds.