planlisted

# Plan YOU do the decomposition — read the code first. `taskops_plan` only makes it durable. **Understand the work before you split it.** One `megabrain_ask` or a read of the real files. A plan written from the user's sentence alone invents tasks that do not match the code. Then call `taskops_plan` ONCE with the whole tree. Each entry: - **`title`** — the outcome, not the activity. "Bind commits to tasks", not "work on commits". - **`spec`** — the brief for a FRESH agent with none of your context, possibly on another developer's machine three days from now. It must say: what **done** looks like concretely enough to disagree with, what must **not** change, and **where to start** (the files, the sibling to copy, the test that pins the behaviour). A one-line spec is the single most common cause of an agent doing the wrong thing correctly. - **`files`** — the edit surface. This is what stops two agents being handed the same file. - **`after`** — dependencies, as the 0-based index of an earlier entry in this same call. ```json [ {"title": "Add the events table", "spec": "���", "files": ["storage/_events.py"]}, {"title": "Board reads it", "spec": "…", "files": ["render/board.py"], "after": [0]} ] ``` **Check the last line of the output.** If it says NOTHING is ready, your `after` references have a cycle or an off-by-one — fix it now, because otherwise the first agent to ask for work gets nothing and nobody will know why. Prefer fewer, larger tasks over many