A portable, self-improving, multi-IDE agentic office: canonical artifacts compiled into Claude, Codex, and Cursor.
This repository also configures its own agents. See what cohort tells them →
Command Claude Code
Recurring deep adversarial audit of the application and the business that ships it — rotates dimensions and subsystems, keeps a coverage ledger, and sweeps the critical path and go-to-market every run.
Command Claude Code
Explain something Barney style — so simple, so step-by-step, that nobody can get it wrong.
Command Claude Code
Recurring generative sweep for new opportunities and ideas — the creative sibling of /audit. Diverges wide across creative lenses, converges on the strongest, turns survivors into pursuit plans, and keeps an idea ledger so it compounds run over run and never re-generates a dead idea.
Command Claude Code
Implement the next task incrementally — build, test, verify, commit.
Command Claude Code
Conduct a five-axis code review — correctness, readability, architecture, security, performance.
Command Claude Code
Simplify code for clarity and maintainability — reduce complexity without changing behavior.
Command Claude Code
Get a second opinion from ChatGPT (via the OpenAI Codex CLI, read-only) on a hard problem — advisory, cross-examined, never executed blindly.
Command Claude Code
Get a second opinion from Grok (via xAI's API, direct) on a hard problem — advisory, cross-examined, never executed blindly.
Command Claude Code
Cross-vendor build — a coordinator routes work to the cheapest capable model from any vendor; Claude subagents write, external engines propose gated patches.
Command Claude Code
Rate an office agent or command from inside the IDE (fuels the improvement loop).
Command Claude Code
Drive a GitHub issue to a draft PR — build, then an independent judge verifies each criterion (max 3 rounds).
Command Claude Code
Interview the user and tailor the global office — wiring, context, and custom agents.
Command Claude Code
Break work into small verifiable tasks with acceptance criteria and dependency ordering.
Command Claude Code
Interview the team about this repo — fill project context and scaffold tailored specialists.
Command Claude Code
Metric-gated autonomous optimization loop — propose, evaluate, keep the gain or revert, in a throwaway worktree; you review the staircase and merge.
Command Claude Code
Multi-vendor adversarial review — fan disjoint areas to Claude tiers, ChatGPT, and Grok, cross-examine over two rounds, synthesize with convergence.
Command Claude Code
Run the pre-launch checklist via parallel fan-out to specialist personas, then synthesize a go/no-go decision.
Command Claude Code
Capture a session snapshot for this repo's Cohort context, filled from this conversation.
Command Claude Code
Start spec-driven development — write a structured specification before writing code.
Command Claude Code
Run TDD workflow — write failing tests, implement, verify. For bugs, use the Prove-It pattern.
Command Claude Code
Update Cohort to the latest upstream version and recompile your IDEs.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: