>
← Agent workflow timeline
2026 February
Agent workflow stage

From repo sessions to a vault-backed operating layer.

How Claude Code and life-os-x connected coding tasks, notes, schemas, and agent context into a markdown-first knowledge and action layer.

Markdown vaultsystem/ contextschema thinkingClaude Code
Before

Repo context and agent context were mostly task-local.

After

Agent work had a durable knowledge substrate outside the coding session.

Anatomy
01

Input

Vault notes, system folders, task records, repo work, agent context.

02

Shift

The source of truth moved out of chat and into durable markdown structures.

03

Output

A vault-backed way to preserve context, route work, and connect coding with personal/project knowledge.

04

Limit

The system was powerful but internally named and still evolving. It needed better routing, public language, and procedural reuse.

Why this mattered

The missing layer was a command bus plus reusable skills that could make this operational day-to-day.

Stage interpretation
What this stage proves

Vault-backed context made agent work less disposable. Notes, tasks, and project context could survive beyond a single CLI session.

What it did not solve

Durable context alone did not guarantee good routing, proof, or public explanations. The next stage needed command-bus behavior and explicit procedural memory.

Page role

Page role: this page explains why durable context mattered. It is not the Personal Knowledge System component map; that current system lives on its own page.