For knowledge or decisions that should survive beyond a chat, app, or model session.
Methodology / Operating model
Vault-first Operating Model
A framework for keeping Markdown, schema, validation, provenance, and human-readable context as the durable source of truth.
Markdown→System→Schema→Validate→Provenance→Use
How it works
Capture in Markdown, structure through system folders and schema.yaml, validate before trusting, and keep caches clearly downstream of the vault.
This overlaps with Systems, but the methodology is the decision rule: source-of-truth first, cache/tool second.
Operating guide
Best for knowledge that should outlive a chat or tool cache: operating rules, project context, decisions, captures, recurring workflows, and source-backed notes.
Write durable context to the canonical vault location, attach provenance, validate schema/links where possible, treat generated indexes as downstream, and report changed paths.
A Telegram capture becomes a scoped vault note or queue item with source, chat/thread/session IDs, project slug when known, and validation evidence.
The model fails when agent memory becomes the real source of truth, loose inbox notes bypass schema, or automation silently drops failed vault writes.
What you get
The minimum output is a durable vault record or validated no-write event with provenance, changed paths, and downstream index/cache status. If the vault cannot be written safely, report the failure visibly.
Taxonomy boundary
Related shelf: `/personal-knowledge-system/` maps the system components; this page is the operating rule that decides where truth should live.