Trove spec work

Trove turns described work into running work. You author specifications in an OPAL conversation — they live in a database, not in files you have to keep tidy — and lanes pick up the resulting work units and run them autonomously, each in its own worktree.

These guides cover both halves: describing the work, watching it run, and choosing which model each role uses.

Guides

• Create a spec — author and refine by conversation; the spec.* tasks do the writing, and every unit carries a Definition of Done.
• Work on several things at once — break work into units, let the lanes run them in parallel, and read progress off the board.
• Run a plan — hand work units to lanes, watch progress, regroup what did not land.
• Work a side quest — work that belongs to no spec, filed and run without inventing a home for it.
• Choose models per role — set the audit / writing / implementation models globally and override them per project.

Related

• Trove — autonomous coding in lanes — the base system: lanes, sessions, and the autonomous implementation cycle.
• Lanes, and why there is only one kind — what the retired spec/impl worktree pair was for, and what replaced it.
• The spec database — why the specification stopped being a file, and what that bought.
• Spec database OPAL command reference — every task, and the rule each one enforces.