Skip to content

Repo Atlas

Top-Level Surfaces

Surface Role Current status
docs/ approved human-readable canon active
research/ pre-canon working models and drafts active
records/ decisions and proposals active
meta/ machine-readable control metadata active
governance/ standards and lifecycle rules active
templates/ reusable governed artifacts active
specs/ versioned contracts and registries active
evidence/ proof of execution and closure active
archive/ historical or retired material reference only
adr/, rfc/ transition pointers compatibility only

Structural Map

flowchart LR
    A["Ledger"] --> B["docs/"]
    A --> C["research/"]
    A --> D["records/"]
    A --> E["meta/"]
    A --> F["governance/"]
    A --> G["specs/"]
    A --> H["templates/"]
    A --> I["evidence/"]
    A --> J["archive/"]

    B --> B1["operating-system/"]
    B --> B2["strategy/"]
    B --> B3["product/"]
    B --> B4["architecture/"]
    B --> B5["company/"]

    C --> C1["operating-system/"]
    C --> C2["strategy/"]
    C --> C3["product/"]
    C --> C4["company/"]

    D --> D1["decisions/"]
    D --> D2["proposals/"]

Reading Order

  1. Start with README.md
  2. Use Canon Indexes from this site to find the right approved entrypoint
  3. Use Research Indexes when canon does not yet answer the question
  4. Use Records And Control Surfaces when the task depends on history, contracts, metadata, templates, or evidence
  5. Move into docs/agents/ and docs/operating-system/ for task execution behavior
  6. Read docs/strategy/ and docs/product/ for current business and shell truth
  7. Read research/strategy/ only when you need long-term architecture reference

Important Structural Rule

Ledger is a control repo. It is not the place for runtime application implementation.

That boundary is what keeps the repo navigable:

  • control truth stays here
  • product/runtime code stays in sibling repos

Stronger Sources

Use the repo-owned family roots when the atlas summary is not precise enough: