Working ledger for ebuild. The planner writes entries; each owning role
updates its own row. Roles are in AGENTS.md, the workflow in
ORCHESTRATION.md, the gate in VERIFY.md.
Status is one of: todo, in-progress, blocked, review, done.
| ID | Task | Owner | Mode | Status | Depends on |
|---|---|---|---|---|---|
| — | No active tasks. | — | — | — | — |
| ID | Task | Owner | Verified by | Evidence |
|---|---|---|---|---|
| — | None yet. | — | — | — |
### T-000 — <short title>
Owner: <role>
Mode: <see MODES.md>
Status: todo
Depends on: <task ids, or none>
Goal
: <one sentence: what is true afterwards that is not true now>
Acceptance criteria
: - <observable, checkable statement>
- <observable, checkable statement>
Files in scope
: <paths the owner is expected to touch>
Out of scope
: <what this task deliberately does not change>
Risks
: <what could break, and what would reveal it>
Verification
: | Check | Command | Result |
|-------|---------|--------|
| <name> | `<command>` | `NOT RUN` |These commands were derived from the manifests at the repository root. Confirm one works before relying on it; a listed script may still be a stub.
| Check | Command | Default state |
|---|---|---|
| Unit tests | pytest |
NOT RUN |
| Build | cmake --build build -j |
NOT RUN |
- One task per unit of work that can be verified on its own.
- Acceptance criteria are written before work starts and are not edited to match what was built. If they were wrong, say so and rewrite them explicitly.
- A task reaches
doneonly when the definition of done in ORCHESTRATION.md is met and the verification commands were actually run. blockedrequires a note naming what it is blocked on and who can unblock it.