Markdown

The nine document families use a YAML frontmatter namespace:

---
langdrift:
  type: prd
  schema_version: 1
  id: PRD-001
  title: Product Vision dashboard
  owners: [user_42]
  research: [RES-002]
  vision_targets: [VT-001]
---
 
# Product Vision dashboard
 
Rich Markdown context remains in the body.

The parsed payload is:

{
  metadata: ArtifactMetadata
  body: string
}

type, schema_version, id, title, and owners are required. IDs are nonempty strings without whitespace; owners are identifiers rather than embedded user profiles.

Optional structured metadata includes:

• description, summary, status, scope, and tags
• deciders, questions, and timestamps
• typed relationship arrays such as research, prds, specs, and vision_targets
• general references, supersedes, and superseded_by
• an external provider identity

Unknown fields inside langdrift are rejected. Root-level YAML outside the langdrift namespace is not protocol metadata and is ignored by parseArtifact.

Parser rules

The frontmatter delimiter must begin the file, apart from an optional UTF-8 byte-order mark. The parser rejects malformed YAML, custom tags, duplicate keys, aliases, and schema-invalid metadata.

The body starts immediately after the closing delimiter. Unicode and original line endings are retained, and body text is not executed. Headings, prose, and filenames do not supply identity or fill missing metadata. Relationships are also taken only from structured metadata.

Record and Drift Event envelopes do not use this Markdown wrapper. They are validated directly from JSON.