Superpowers: make a coding agent follow a complete development methodology

Who it's for

Look at Superpowers when you need a complete software-development methodology for a coding agent. The official definition is: Superpowers is a complete software development methodology for your coding agents, built on top of a set of composable skills and some initial instructions that make sure your agent uses them.

It relies on skills triggering automatically; you are not asked to do a special summon. It fits people who already have a coding agent and will follow “sign off the design first, then produce an implementation plan.” If you only want a one-shot command and will not accept that methodological pace, it is a mismatch.

Page: https://github.com/obra/superpowers

First try in 3 minutes

Install steps differ by harness. For the first try, use only this official command:

/plugin install superpowers@claude-plugins-official

Preconditions: this harness can run the plugin command above. Official wording: Installation differs by harness. If you use more than one, install Superpowers separately for each one. No on-disk directory is written; do not invent a path.

After install, do not change code at once. First ask the agent to confirm that skills trigger automatically. The success signal is that the agent can restate And because the skills trigger automatically, you don't need to do anything special.

Then give a design you have already signed off (read-only; do not let it start writing) and ask it to produce only an implementation plan. The success signal is a plan clear enough that an enthusiastic junior engineer with poor taste, no judgement, no project context, and an aversion to testing could follow it, and the repository has not been changed yet. If the agent skips the plan and starts touching files, stop.

Use cases

Case 1: Produce an implementation plan after the design is signed off. The body says: After you've signed off on the design, your agent puts together an implementation plan.... Hand the agent a design already agreed, and ask for a plan only, not implementation. What you can check is that the plan is detailed enough for a junior engineer who lacks judgement, has no project context, and dislikes testing. The page gives no plan-template filename; do not invent one.

Case 2: After install, work by automatic triggering. The body says you don't need to do anything special. Your coding agent just has Superpowers. Fits watching, after install, whether a normal coding conversation applies the methodology by itself, rather than calling a skill name that was never written. If it only works when you name it by hand, install on this harness has not taken effect; if you use more than one harness, install separately on each.

Usage limits

The same page also has /add-plugin superpowers. That is another harness’s install wording, not the main path for the first try; do not promote either command to every environment. Multiple harnesses must be installed separately.

The methodology sits on composable skills and initial instructions. The page does not enumerate the skills or name clients; do not add names, and do not write it as supporting only one agent.

Design sign-off comes first, then the implementation plan. The page does not authorize starting work before sign-off.

No secrets, uninstall command, or local directory. Do not invent unwritten test steps, branch strategy, or publish actions.