From documents to a task plan.

Documents are the living corpus. Plan it snapshots them into an epic spec. Then the breakdown produces stories an agent can finish.

Write in Documents first

Do not start by typing a spec into an empty epic. Write the PRD and supporting pages. The richer the corpus, the better the snapshot.

# Product requirements

What are we building, and why? The agent can co-write this page. It cannot silently overwrite you.

## Problem

Agents invent scope when the work was never decided.

Plan it, then break it down

Select pages, compose the epic spec, optionally run breakdown. Each story gets a title, description, acceptance criteria, type (feature / bug / chore), and Fibonacci points (1, 2, 3, 5, 8). You import. The agent does not.

Plan it

Snapshot selected documents into an epic spec. The epic stays independently editable after this.

prd

research/auth

notes/parking-lot

Destination: backlog · Breakdown: on

Stories sized for a context window

1–5 points stay inside a typical agent window. 8 is the intentional stretch. Combined with acceptance criteria, “done” is a list, not a vibe. There is no velocity dashboard. Completed-panel point totals are just counts.

From that snapshot

Bring your own keys

Breaking down a spec costs 10 credits on Beck’s AI. Expanding it costs 5. A question costs 2. Add an Anthropic or OpenAI key on any plan, including Free, and those calls skip credits and daily limits.

Don’t spend the month counting credits.

A breakdown is 10. Free is 50. If this is how you plan, Pro is $12.

See Plans

Free is 50 credits. Enough to see if the loop is real.