Skip to content

Subagent-Driven Development skill

Execute an approved plan with fresh implementers, scoped reviews and a recorded task ledger.

Community · Jesse Vincent / Prime Radiant
AuthorJesse Vincent / Prime Radiant
LicenceMIT
Sourceobra/superpowers · SKILL.md on GitHub
Project websiteprimeradiant.com/superpowers
Works inClaude Code
Read from the source on

The subagent-driven-development skill executes a prepared plan through a controller and fresh task implementers. Each task receives a focused brief, followed by specification and code-quality review.

It maintains a plan-specific ledger and review packages, then performs a whole-branch review. This requires real subagent and repository capabilities; a text-only chat cannot reproduce the complete workflow.

From the skill's SKILL.md: subagent-driven-development

This is a community skill by Jesse Vincent / Prime Radiant, not checked by us or by Anthropic. Skills can tell the agent to run code, read files or call the network. Before installing, read its SKILL.md and every script in the folder, and look for anything that does not match what it says it does.

What it does

  1. Prepare the workspace, task ledger and bounded implementation briefs.
  2. Run one task implementer at a time, review its result and manage scoped fixes.
  3. Review the whole branch and report unresolved findings and recorded rulings.

What you get

  • Implemented changes, verification reports, review packages and a plan-specific .superpowers/sdd ledger.

When to use it

  • Execute a reviewed plan whose tasks have clear interfaces and completion checks.

When not to use it

  • A host without subagents, shell tools or repository access.

What it needs

  • A subagent-capable coding host, Git, shell tools and the complete companion scripts.
  • Additional agents consume model usage. Tests and dependencies remain project-specific.

Files in the skill

File or folderWhat it is for
SKILL.mdController workflow and model-selection boundaries.
implementer-prompt.mdFocused task brief for the implementer.
task-reviewer-prompt.mdCombined specification and quality review.
scripts/sdd-workspacePlan-specific workspace and ledger helper.
scripts/review-packageBounded diff package for reviewers.
references/fix-loop.mdLimited fix rounds and adjudication rules.

Licence as found in the repository on 2026-11-19: MIT.

Example prompts

Type one of these after installing; naming the skill makes sure it is used.

  • Execute this approved plan with one fresh implementer per task and preserve the review ledger.
  • Review the task’s specification and code quality using a scoped diff package.
  • Report every recorded ruling and its risk before finishing this branch.

How to install subagent-driven-development

Installation and permissions

Use the complete Superpowers plugin. It references worktree, testing, code-review and branch-finishing skills outside this individual folder.

Review the bundled shell helpers first. The workflow can create branches, run tests and make commits; it must respect the host’s permissions and repository instructions.

A skill is a folder with a SKILL.md file. You install it once; the agent reads its short description at the start of each session and loads the full instructions only when a request matches. Pick the place you use:

Claude Code: from the plugin marketplace

  1. Start Claude Code in a terminal with claude.
  2. This plugin is in Anthropic's official marketplace, which Claude Code adds for you the first time you start it.
  3. Install the plugin. Claude Code opens its details first: check what it will add (skills, hooks, MCP servers), then choose whether it is for you, for everyone on this repository, or for you in this repository only.
    /plugin install superpowers@claude-plugins-official
  4. Type / and look for /superpowers:subagent-driven-development, or simply ask for the task.

From your shell instead of a session:

claude plugin install superpowers@claude-plugins-official

Claude Code: copy the folder into your skills folder

Put the skill's folder in ~/.claude/skills/ to use it in every project on this computer, or in .claude/skills/ inside a repository to share it with everyone who works there (commit it).

git clone --depth 1 https://github.com/obra/superpowers.git
mkdir -p ~/.claude/skills
cp -R superpowers/skills/subagent-driven-development ~/.claude/skills/

Claude Code notices new or changed skills in these folders during the session. If the skills folder did not exist when the session started, run /reload-skills once. Then type /subagent-driven-development or just describe the task.

On Windows, run the commands in Git Bash or WSL, or copy the folder into the same place with File Explorer.

Only install skills you trust. Anthropic's advice is to use skills only from trusted sources, and to read every file of any other skill before using it: a skill can direct the agent to run commands, read files or send data in ways its description does not mention.

Sources for these steps

Checked against each vendor's documentation on 2026-11-14:

Keep decisions visible

Task briefs and diff packages keep reviewers focused without inheriting the whole controller conversation. The ledger records deferred findings and decisions.

A requirement that cannot be verified from the diff still needs the controller’s investigation. It is not automatically a passing check.

Questions about subagent-driven-development

Does it run implementers in parallel?

The described task loop uses one implementer at a time. Fresh context does not mean concurrent implementation.

Can fixes loop forever?

The source bounds task fix rounds and records adjudications. A final review has one fix wave and one scoped re-review.

Does review prove there are no defects?

No. Reviews and tests provide bounded evidence; unresolved findings and rulings must remain visible.

Will it work in a hosted chat or sandbox API?

The complete workflow needs subagents, Git, shell execution and repository access. Hosted app/API compatibility is not established here.

More AI skills

Related skills

Skill categories