Claude Code Plugin
v1.0.0

superjawn

Skills that research before they act.

A fork of obra/superpowers. Each entry-point skill runs research by default — before brainstorming, before debugging, before writing skills. Stale-artifact consumers get a narrow freshness check. Everything else ports clean.

What this is

superjawn is a research-augmented variant of obra/superpowers (MIT, Jesse Vincent, v5.0.7). The upstream plugin ships 14 skills covering the full development workflow — brainstorming through finishing a branch. superjawn forks those skills and adds structured research at entry-point stages.

The idea: when you start new work without an upstream artifact (a spec, a plan, a test), a brief research phase catches pitfalls, surfaces prior art, and checks your memory before you commit to an approach. When you're consuming an artifact that was already research-baked, re-researching wastes time and risks conflicting conclusions.

So research fires in three places only: brainstorming, systematic-debugging, and writing-skills. Stale-artifact skills get a lighter freshness check. Everything else is a straight port.

Forked from obra/superpowers

MIT-licensed. Synced against upstream v5.0.7. Modifications carry the same license. Copyright 2025 Jesse Vincent; modifications copyright 2026 Joe Amditis.

Three categories, 14 skills total

Research phase (3 skills), freshness check (1 skill), consumer/straight port (10 skills). v1.0.0 ships all 14 with no soft dependencies on the upstream superpowers plugin — code-reviewer agent dispatches now target pr-review-toolkit:code-reviewer.

Coexists with upstream

Keep both superpowers and superjawn installed during rollout. Skills not yet ported resolve via superpowers:<skill>. Upstream disabled at v1.0.0 after all 14 skills land and prove out.

How research works

Three categories, one principle: research belongs where work starts from scratch, not where it inherits a baked-in conclusion.

Research

3 entry-point skills

brainstorming, systematic-debugging, writing-skills

Research phase fires by default. Skip only with explicit, written justification per the locked skip protocol.

Web — error string searches, GitHub issues, current best practices
Codebase prior-bugs git log --grep for prior failures in this area
Authoritative — current docs for any external APIs in scope
User-context — memory check for relevant prior decisions
dispatch varies per skill — see each skill's research-phase section
Freshness check

1 stale-artifact skill

executing-plans

Default-skip — the inverse of research. Fires only when a trigger indicates real drift risk.

Cross-session execution — plan drafted in a prior session
External API touched — any task calls a live API or service
main/master branch — heightened drift risk on integration branches
Findings: .superpowers/exec-journal-<slug>.md
Consumer

10 downstream skills

writing-plans, test-driven-development, + 8 more

No research phase. These skills consume an artifact from an already-researched upstream stage. The handoff carries the conclusions.

Attribution comment added
Cross-references rewritten to superjawn:
No flow-diagram edits required
Trust the upstream-artifact handoff

What's ported (v1.0.0)

14 of 14 complete
Standalone — no soft dependency on upstream superpowers.
Skill Category Status
brainstorming Research Ported (Batch 1)
writing-plans Consumer Ported (Batch 1)
executing-plans Freshness check Ported (Batch 1)
systematic-debugging Research Ported (Batch 2)
test-driven-development Consumer Ported (Batch 2)
verification-before-completion Consumer Ported (Batch 2)
receiving-code-review Consumer Ported (Batch 3)
requesting-code-review Consumer Ported (Batch 3)
subagent-driven-development Consumer Ported (Batch 4)
dispatching-parallel-agents Consumer Ported (Batch 4)
using-git-worktrees Consumer Ported (Batch 4)
finishing-a-development-branch Consumer Ported (Batch 5)
using-superjawn
renamed from upstream using-superpowers
Consumer Ported (Batch 5)
writing-skills Research Ported (Batch 5)

How to install

You need Claude Code installed. Run claude --version to check.

1

Add the plugin source

Run this once to register the marketplace.

claude plugin marketplace add https://github.com/jamditis/claude-skills-journalism
2

Install superjawn

Install from the marketplace.

claude plugin install superjawn

After installing, restart Claude Code so the new skills load. Existing sessions can't pick up new plugin versions mid-flight — superjawn:* invocations from a session that started before the install will fall through to upstream.

!

Existing superpowers users

v1.0.0 ships with no soft dependencies on the upstream superpowers plugin — code-reviewer agent dispatches now target pr-review-toolkit:code-reviewer (Anthropic-maintained, in @claude-code-plugins). You can run superjawn standalone from this version forward.

If you want to revert to the upstream plugin: /plugin install superpowers@claude-plugins-official.

Update an existing install

claude plugin update superjawn

After installing or updating, restart Claude Code so the new skills load. Existing sessions can't pick up new plugin versions mid-flight — superjawn:* invocations from a session that started before the update will fall through to upstream.

Skip protocol

The research phase is default-on. Skipping is allowed, but requires a written justification. Below is the locked skip-protocol text shared by all three research-category skills (brainstorming, systematic-debugging, writing-skills) — byte-identical across the set so behavior stays predictable across entry points.

brainstorming/SKILL.md — Skip protocol (canonical)
Quoted directly from superjawn/skills/brainstorming/SKILL.md (and byte-equivalent in systematic-debugging/SKILL.md).
If skipping, write one line into the spec doc: `Skipped research because <reason>. <Verifiable pointer if applicable>.`

**Valid reasons:**
- Trivial scope (typo, comment edit, single-line config)
- Fresh prior research — same topic in current session OR within last 7 days with verifiable spec/plan pointer. **If the pointer doesn't resolve, the skip is invalid.** (Beyond 7 days, repeat the research even if you remember the prior findings — the landscape drifts.)
- User explicit — **must quote the phrase** that authorized the skip.
- Repeat of identical task — **must include a pointer** to the prior successful run.

**Invalid reasons:** "I think I know", "seems straightforward", "moving fast", "user wants this done quickly", "already familiar with this codebase". If those are tempting, do the research.

Credits

obra/superpowers

Original 14-skill plugin by Jesse Vincent. MIT-licensed, v5.0.7 baseline. superjawn forks this work with modifications for research phase insertion.

Copyright 2025 Jesse Vincent

Modifications

Research phase design, freshness check logic, architecture revision, and all per-skill modifications by Joe Amditis. Carries same MIT license.

Copyright 2026 Joe Amditis

MIT License — see superjawn/LICENSE for the full text.