Edit: github-backup/deployment/vault-template/SCHEMA.md

Note Content (Markdown)

# Easier Brain Schema

## Domain

Operator knowledge for Easier Agency, Easier Now and approved related business
work: research, decisions, experiments, playbooks and relationship context.

## Status

This is an empty preparation vault. Do not ingest private or customer
information until backup, privacy and access policies have been approved.

## Structure

- `inbox/`: captures awaiting classification.
- `raw/`: approved small source extracts. Bulky raw artifacts stay in source
  systems by default.
- `receipts/`: source IDs, links, owners, capture dates, sensitivity and
  digest status.
- `digests/`: meeting, Slack, Drive, campaign and operating summaries awaiting
  review or promotion.
- `people/` and `organisations/`: relationship context with careful sourcing.
- `projects/`: active initiatives and product work.
- `decisions/`: dated choices with rationale and reversal conditions.
- `experiments/`: hypotheses, tests, evidence and outcomes.
- `playbooks/`: reviewed operating procedures.
- `concepts/`: reusable insights and frameworks.
- `briefs/`: draft or reviewed output assembled from knowledge.
- `reviews/`: periodic audit and synthesis records.
- `agents/`: role-specific agent instructions.
- `sops/`: recurring procedures and role playbooks.
- `heartbeats/`: recurring loops, manual until approved.
- `evals/`: good and bad example outputs for each agent.
- `decisions/`, `concepts/` and `projects/` may include seed pages approved
  for dry-runs before live connectors exist.
- `_archive/`: superseded compiled pages retained for traceability.

## Compiled Page Frontmatter

```yaml
---
title: "Page title"
type: concept
status: draft
created: YYYY-MM-DD
updated: YYYY-MM-DD
entities: []
domains: []
sources: []
sensitivity: public
confidence: medium
review_by: YYYY-MM-DD
contested: false
---
```

Allowed `type` values: `person`, `organisation`, `project`, `decision`,
`experiment`, `playbook`, `concept`, `brief`, `review`, `receipt`, `digest`.

Allowed `sensitivity` values: `public`, `internal`, `client-confidential`,
`personal-sensitive`.

Allowed `confidence` values: `low`, `medium`, `high`.

## Source and Provenance Rules

- Raw evidence is not rewritten after capture.
- Large Fathom transcripts, Slack histories and Drive creative assets should
  usually remain in source systems with a receipt and digest.
- Raw notes should include origin, capture date and a checksum when imported.
- Compiled factual statements cite a source in `sources` and preferably close
  to the claim in the body.
- Agent-created inference is labelled explicitly.
- Contradictions are preserved and set `contested: true`.

## Autonomy Rules

- No external side effects from vault content.
- No agent-created relationship claims treated as facts without evidence.
- No bulk update, sensitive import or data export without human approval.
- No bulk raw artifact import without a digest and retention plan.
- A later approved maintenance job must produce a readable report and diff.