# Brain Consolidate

> Audit, lint, migrate, deduplicate, repair, archive, forget, or delete the configured brain. Use for explicit vault cleanup/doctor requests and clear structural debt. Audit is read-only by default; repairs are planned and destructive or semantic-loss actions require confirmation. Do not use for ordinary capture, targeted recall, generic code consolidation, semantic pattern promotion, or onboarding.

- Skill: `riggyz/brain-consolidate` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add riggyz/brain-consolidate`
- Raw SKILL.md: https://api.skillmd.com/api/skills/riggyz/brain-consolidate/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Security
- Author: riggyz (https://skillmd.com/u/riggyz)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/riggyz/brain-consolidate

---


# Brain Consolidate

Own structural correctness and lifecycle maintenance. Do not invent semantic lessons.

## Foundation Handoff

Invoke `brain` through the host's native skill loader and use the exact base directory it returns for foundation resources and scripts. Do not list or search parent skill directories, read `SKILL.md` directly, or probe alternate config locations.

## Modes

- **Audit:** run the read-only doctor and report findings.
- **Repair:** fix non-destructive routers, links, metadata, tags, relationships, and malformed records.
- **Migrate:** execute an approved architecture-version transaction with backup, ledger, and postflight checks.
- **Archive/forget/delete:** preserve current truth and evidence according to user intent; irreversible loss requires confirmation.

## Audit

Using the exact foundation base directory returned by the skill loader, run:

```sh
BRAIN_FOUNDATION_DIR="<exact base directory returned by the brain skill loader>"
npx tsx "$BRAIN_FOUNDATION_DIR/scripts/audit-vault.ts" --vault-path "<vault-path>"
```

Check root routers, node registration/tags, atomic record reachability/provenance, workspace membership, project dependency reciprocity, cold-history isolation, links, note sizes, backlog lifecycle, issue tags, and managed graph groups. Orphans alone never prove health.

## Plan And Apply

Classify each operation as link, metadata repair, split, move, supersede, archive, soft-forget, graph change, or delete. Surface the plan before semantic-loss operations. Preserve snapshots before large migrations.

Prefer Obsidian-aware moves while the app is running. Treat live app-owned files and concurrent agents as competing writers: reread immediately before mutation and never parallelize a vault transaction.

Non-destructive structural repairs may proceed without per-item approval. Deleting notes/artifacts, merging then deleting a duplicate, discarding unpreserved history, and unrelated graph-group deletion require confirmation.

Every graph mutation follows the foundation graph-maintenance ceremony. If Graph views are open or the reload cannot complete, defer graph work rather than claiming success.

## V1 To V2 Migration

Recognize legacy root typed notes, singular `wiki/`, project-classified workspaces, and flat owner mega-notes as migration inputs. Migrate only after the v2 source suite/doctor is installed or available, a recoverable vault backup exists, source-to-destination coverage is complete, and one writer owns the transaction.

Verify routers, owner tags, links, memberships, dependencies, config paths, graph JSON, unresolved links, orphans, and representative recall after migration.

## Boundaries

- Do not capture ordinary facts or run checkpoints.
- Do not promote repeated patterns or rewrite global semantic records based on frequency.
- Do not mark a backlog item fixed without verifying the defect.
- Product work remains in its tracker.
- Explicit no-write and no-delete instructions win.