Provides deep domain expertise and best practice implementation details. Use to evaluate the initial plan.
Identifies risks and assumptions in the plan. Use to evaluate the initial plan and the architects feedback.
Write production code that implements the approved plan and makes tests pass. The only agent that is allowed to make changes to the domain logic.
Coordinate Power Trio workflows and subagents.
Reviews implementation for issues that negatively impact performance. Use proactively when the feature affects user experience, or lies on the applications critical path.
Reference documentation for the Power Trio voting system.
Reference documentation for Power Trio artifact formats.
Execute the Power Trio build phase only. Use proactively when writing, modifying, fixing, or testing code.
Code is read much more often than it is written. Optimize for the reader.
Centralized configuration for the Power Trio workflow. Reference this file for consistent settings. These values must be adhered to.
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Good things come in threes: meals a day, wishes, and the John Mayer Trio.
A committee of agents for planning, building, and reviewing.
Power Trio is a three-phase AI multi-agent system designed to deliver code quickly but maintains human-quality standards. Each phase operates as a committee of three specialized agents who iterate toward consensus before passing work to the next phase.
The core insight: agents are fast, but the criticism of AI-generated code is "slop"—fast output that lacks quality. Power Trio addresses this by separating speed (Phase 2) from quality assurance (Phase 3), with clear planning upfront (Phase 1).
This is a constraint pipeline, not a reasoning pipeline. Each phase narrows degrees of freedom, removes ambiguity, and converts opinions into enforceable rules.
Add the marketplace:
/plugin marketplace add thesimonho/artificial-jellybeans
Then install the plugin:
/plugin install power-trio@thesimonho/artificial-jellybeans
Run the full workflow:
/power-trio:start <description>
Run the workflow without the review phase:
/power-trio:start --no-review <description>
Or run an individual phases as needed:
/power-trio:plan <description>
/power-trio:build <description>
/power-trio:review <description>
Advanced users might ask: "Why can't I just do Plan → Code → Review with good prompts?"
Power Trio enables what sequential prompting cannot:
Parallel disagreement surfaces issues earlier - Multiple agents with different concerns challenge each other in real-time, rather than sequential review catching issues after the fact
Hard role constraints prevent "model collapse" - Agents can't drift into agreeable slop because their enforcement mechanisms are explicit and bounded
Forced delegation loops prevent patchy fixes - Phase 3 cannot hack in quick fixes; they must delegate back to Phase 2's full committee with proper TDD and review
Repeated committee patterns reduce prompt fiddling - The structure is consistent and predictable, not dependent on crafting perfect sequential prompts
See PHASES.md for the detailed description of each phase.
See ORCHESTRATION.md for the orchestration logic within and between each phase.
Goal: Reach consensus on what to build and how to build it
Committee:
Why This Committee Split Works:
Goal: Implement the plan quickly without sacrificing long-term code quality
Committee:
Why This Committee Split Works:
Goal: Ensure production-readiness and validate the solution solves the original problem
Phase 3 is optional - automatically triggered when risk heuristics are met.
Committee:
Why This Committee Split Works:
do [do]cumentation.
npx claudepluginhub thesimonho/artificial-jellybeans --plugin power-trioDescribe your goal, approve the spec, then step away — Claude and Codex loop together until it's right.
Persona-driven AI development team: orchestrator, team agents, review agents, skills, slash commands, and advisory hooks for Claude Code
Plan and autonomously build a software task end-to-end. Recons the codebase, applies preloaded memory, decomposes into the right number of phases, gets one confirmation, then prepares a single ready-to-paste /goal command — one paste between you and done — that drives execution to completion with built-in retry, fix-spec recovery, and per-phase memory writeback. Works on Claude Code and Codex.
Structured project planning and execution through brainstorm, spec, and build phases across three execution tiers: sequential, delegated sub-agents, and full agent teams
Multi-agent software evolution. Autonomously study, strategize, build, review, and evaluate code changes using specialist AI agents.
Claude harness - A harness for solo developers (Vibecoders) to handle full-cycle contract development.