rota

rota vs Superpowers

Superpowers is “a complete software development methodology for your coding agents, built on top of a set of composable skills.” It is a skills library: brainstorming, writing-plans, subagent-driven-development, test-driven-development, systematic-debugging, code review and a few more, which trigger on their own as you work. Checked 2026-10-26 against Superpowers v7.0.1 (released 2026-10-21).

The short version

  • Pick Superpowers when you want a disciplined way of working inside one session, on many agent harnesses, with nothing to install but skills.
  • Pick rota when work comes from a backlog, several issues should run side by side, and a merge gate should verify the result before anything lands on main.
  • Both are skills, so they overlap in the solo loop (design, plan, build, review). They differ most once you have more than one agent.

Side by side

  Superpowers rota
What it is Skills methodology Single Go binary plus /rota-* skills
Unit of work A feature or task in the current conversation, broken into plan tasks A backlog item or tracker issue
Parallel agents A fresh subagent per plan task, run inside one session; dispatching-parallel-agents for independent problems An orchestrator plus standing workers, one issue each, in separate sessions
Isolation using-git-worktrees sets up a worktree for the work One git worktree per worker, reused across issues
Verification before merge verification-before-completion (run the command, read the output), per-task review, final branch review, full suite before the finish menu Targeted checks per worker, then a gate that runs test.full on a scratch merge before merging
Who merges You pick from a menu: merge locally, open a PR, or keep the branch The orchestrator, through the gate; workers open PRs and never merge
Backlog / tracker Not documented .rota/BACKLOG.md, GitHub, GitLab, Jira, Linear and others
Memory across sessions Plan files in the repo; no persistent memory layer documented KNOWLEDGE.md, DECISIONS.md, handoff notes in .rota/
Agents and models 16 harnesses listed, including Claude Code, Codex, Cursor, Gemini CLI, Copilot CLI, OpenCode Claude Code and Codex; OpenCode workers
Interface Skills inside your agent; no CLI Skills plus the rota CLI, terminal only
Install Plugin install, separately per harness brew install l4ci/tap/rota or the install script, then rota skills install; also a Claude Code plugin
License MIT MIT

Where Superpowers is stronger

  • Reach. It lists 16 harnesses with per-harness install steps. rota supports Claude Code and Codex, with OpenCode as a worker harness.
  • Nothing to run. It is skills only. rota needs a binary on your PATH, a .rota/ folder in the project and, for rounds, gh or glab and a terminal host.
  • Skills fire on their own. using-superpowers makes the agent look for a matching skill before it answers. You do not type a command. rota skills are mostly invoked by name.
  • Within-task discipline. Fresh subagent per task, a spec and quality review after each, and a whole-branch review at the end. subagent-driven-development also tells the agent to record rulings instead of stalling, so a plan can run for a long stretch without you.

Where rota is stronger

  • Several issues at once. A round runs two to five standing workers in parallel, each in its own worktree and terminal tab, with the orchestrator assigning, relaying answers and merging. See parallel rounds. Superpowers’ parallelism is subagents inside one session.
  • A merge gate. rota worker gate runs the full suite on a scratch merge and merges only on a pass, so a red result does not reach main. See how it works.
  • A real backlog. Items live in a file or in GitHub, GitLab, Jira or Linear issues, and workers pick them up by ID. See issue backend.
  • Usage limits. round.accounts spreads workers over several accounts, and rota limit watch moves an issue to a free slot when one hits a limit. See parallel rounds.
  • Durable project memory. /rota-learn and /rota-decide write to KNOWLEDGE.md and DECISIONS.md; /rota-pause leaves a handoff note. See learning and decisions.
  • Recorded proof. /rota-work requires a red run, then a green one, through rota proof record. See running work.

Using both

They are both skills, so you can install both. The combination is untested. The practical question is overlap. Each ships a brainstorming skill, a planning skill, TDD, debugging and review, and both descriptions tell the agent to use them early. If both fire, the agent has two competing playbooks for the same step.

A split that avoids that:

  • Use rota for the backlog, rounds, the gate and project memory.
  • Borrow Superpowers’ within-task skills where rota has no equivalent step, for example verification-before-completion or systematic-debugging, and leave its planning and branch-finishing skills off so they do not compete with /rota-plan and /rota-ship.

finishing-a-development-branch ends by asking you to merge, open a PR or keep the branch. In a round, workers must not merge, so that skill conflicts with the worker contract.

Sources

More comparisons: how rota compares.

Edit this page on GitHub