ulw-loop (Ultrawork Goal Loop)

What it does

ulw-loop enforces a "goal → atomic step → evidence" discipline for long-running delivery.

• Registers goals in the omo ulw-loop CLI state (.omo/ulw-loop) and mirrors every step into a live checklist with exactly one in_progress item.
• Passing tests never count as done. Each success criterion needs observable evidence from a real surface: terminal/TUI, HTTP, browser, computer-use, CLI stdout, DB diff, or parsed config dump.
• Evidence is bound to the git tree hash it was captured at, so it goes stale when tracked content changes and must be re-captured at the current HEAD.
• Code edits, test writing, fixes, and QA are delegated to right-sized subagents via the native task tool, run in the background with WORKING: / BLOCKED: progress reporting and task_send corrections.
• Team mode (team_create) is a deliberate decision, only when work units genuinely overlap — with one git worktree per member and per-unit merges.
• After compaction, it re-reads brief + goals + ledger and omo ulw-loop status --json before resuming instead of re-planning.

Who it's for

• Engineers running multi-day refactors or migrations where context loss is inevitable.
• Teams tired of "tests passed but it didn't actually work" and wanting verifiable receipts.
• Anyone orchestrating many parallel agents who needs file-scope isolation and clean atomic commits.
• Best suited to users already on (or willing to install) the omo-senpi / omo CLI stack.

Examples

1. Large migration: "use ulw to move the auth module to the new SDK" — goals are registered, steps decomposed, and each step captures a real login run in a terminal or browser as evidence.
2. Parallel feature work: batch-spawn three background workers for three features in disjoint files, while the lead reports active subagent count and the latest WORKING: phase.
3. Manual QA checkpoint: before release, collect evidence per success criterion (HTTP response, DB diff) and commit each verified unit atomically in the repo's own message style.

Install the skill I found on Claude Skill Mart.
Copy the packages/omo-senpi/skills/ulw-loop folder from the GitHub repo code-yeongyu/oh-my-openagent into my ~/.claude/skills/code-yeongyu-ulw-loop/.
When it's done, tell me in one line what this skill can do.