Backlog & tasks

Two small files under .atl/ hold a project's deferred decisions and its near-term intentions. They are decision state — a sibling to the knowledge system's journal and wiki, not a third knowledge layer — and they are written and kept current by the /brainstorm skill.

The two tiers

File	What it is	How it's kept
.atl/backlog.md	The passive, trigger-gated superset of everything deferred, punted, or left uncertain. A scannable index, not a to-do list.	Grouped by area; an item is deleted when it ships or is promoted
.atl/tasks.md	The active-intent subset — the short, prioritized list of what you actually mean to do next.	Now / Next; an item is deleted when it ships

An item moves backlog → tasks when you decide to pull it forward — a trigger fired, or you simply chose to prioritize it. That is the whole relationship: the backlog is everything you've consciously chosen not to do now; tasks is the slice you've chosen to do.

backlog.md

The backlog exists so nothing is lost when scope moves on. Every time a brainstorm defers a sub-topic, marks something "not in this step", or leaves a question open, it lands here immediately — so a decision made months ago is still discoverable when its moment comes.

• Grouped by area, not by date. Headings are themes (## Learning loop, ## Distribution, …); the file is an index you scan, not a chronology.
• One line per item: - **Title** — one sentence. _Trigger:_ when it resurfaces. ↳ [source](...).
• Trigger-gated. Most items carry a _Trigger:_ — the condition under which they come back. The backlog is not a to-do list; it's the memory of what you deferred and why it would return.
• The detail lives in the source. The rich "why deferred / full context" stays in the linked brainstorm — the backlog is the index, the brainstorm is the record. Don't duplicate it.

tasks.md

Tasks is the honest short list of active intent.

• Format: - [ ] **Title** — one sentence. ↳ [source](...), grouped under ## Now / ## Next.
• Short and honest. If nothing is actively planned, tasks.md is nearly empty — that is the correct state, not a gap to fill. Don't manufacture tasks; unplanned deferred work belongs in backlog.md.
• Deleted when shipped. A finished task is removed (the docs/ and CLAUDE.md become the source of truth), never left behind as a checked-off item.

Lifecycle

brainstorm defers  →  backlog.md  →(pull forward)→  tasks.md  →(ship)→  deleted
                          └──────────────(ship directly)──────────────→  deleted

• Into the backlog: /brainstorm done scans the closed brainstorm and files every deferred or uncertain item under the right area group. Skipping this is how deferred scope silently disappears.
• Backlog → tasks: promote an item when you decide to act on it.
• Out: an item leaves either file the moment it ships — the docs become the truth, and nothing lingers as done.

Board-backend projects

A project that runs a delivery board backend — one with a .delivery/config.json carrying a backend field, written by /delivery-init — does not use these two files. Its project board is the single authoritative deferral surface (one surface, so the board and the .md files can't drift apart). /brainstorm done detects the config and syncs every deferred and actively-intended item to the board as a work-item — idempotently, keyed on the brainstorm's name + item title before creating so a re-run converges rather than duplicates — and stops writing the .atl/ files. Retiring any content those files already hold (a pointer to the board + a one-time migration) is a separate, per-project step. Everything above this section is the default for every project without a board backend.

Scaffolding

atl init (and atl install) drop empty backlog.md and tasks.md skeletons under .atl/, only if they don't already exist — your own files are never overwritten. The global tier has no project .atl/, so it is skipped.