Fourteen specialized slash commands turn one Claude session into a structured multi-agent engineering pipeline. Decision agents orchestrate, triage, and review; workers execute in parallel — every agent runs on your current session model, every step is a Worker → Reviewer pair, and every non-trivial phase fans into sub-phases with their own reviewers. Auto-routing is on by default — say 'audit the diff', 'debug this test', 'large migration', or 'run a workflow' and the orchestrator routes to the right skill without the /hyperflow:* prefix. /hyperflow:workflow uses Claude Code dynamic workflows for big tasks and a portable Codex/OpenCode adapter where native workflows are unavailable; /hyperflow:plan rewrites a rough prompt into a high-quality one, asks the questions a senior engineer would, and decomposes the approach into a batched task graph; /hyperflow:dispatch fans out persona-stitched workers under per-batch and final-integration review; /hyperflow:design establishes a domain-grounded design system, researches real-world prior art, and renders it slop-free under the local taste skills. 15 composable personas, 6 adaptive flow profiles, and persistent project memory compound across sessions. Works across Codex App/CLI, Claude Code, OpenCode, Antigravity, and Cursor.
Multi-agent orchestration for Codex App/CLI, Claude Code, OpenCode, Antigravity & Cursor.
Decision agents plan and review every step; workers execute in parallel — all on your current session model. Learnings persist in local, per-project memory.
plan → dispatch → audit → deploy
Start anywhere. Auto-advance forward. Memory persists across sessions.
Not just another orchestrator — three things set Hyperflow apart:
Every step is reviewed. Worker → Reviewer is an iron rule at every granularity, sub-phases included. No worker output ships unreviewed.
Memory that's yours. Learnings, decisions, and pitfalls persist in .hyperflow/memory/ — plain markdown, committed with your repo, never uploaded, never mixed across projects. Hot/warm/cold tiering keeps injection cheap.
Depth that adapts. Triage classifies every task and picks a flow profile (fast → scientific), so a 5-line fix never triggers a 300k-token deep run.
Compaction at the right boundary. Automatic context compaction is held until dispatch reaches its end-of-chain gate, then checks estimated transcript usage and snapshots task state before compacting.
Underneath: a structural role separation (decision agents plan & review, workers execute — all on your session model), 15 persona-stitched experts, intent auto-routing, and host-agnostic dispatch — all local, no daemon.
The chain
Start with a rough idea — the pipeline carries it to shipped. Start at any entry point; the orchestrator picks up and runs forward.
#
Skill
What it does
1
plan
Front door — sharpen the prompt (persona standards + 8-dim rubric), design the approach (multi-dimensional analysis + alternatives, refuses to code before you approve), and decompose it into a parallel task graph with a full, test-complete implementation brief per sub-task (incl. an end-to-end case) so the build runs faithfully on a cheaper model; skips whichever phase the request doesn't need
2
dispatch
Fan out persona-stitched workers under per-batch + final-integration review
3
workflow
Big-task lane — native Claude Code workflows, custom Codex/OpenCode adapter
plan never implements — it stops at a build-location gate (build here → dispatch, another session → handoff, or just keep the plan), so you stay in control of when a build starts; audit and deploy are gates that fire at the end. Enter at plan for anything from a rough idea to a clear task — it bounces straight to decomposition when the approach is already clear — or at dispatch when a task file already exists. scaffold is a one-time project setup — run it once per repo to build the .hyperflow/ cache.
workflow is the big-task lane. Hyperflow routes deep/scientific/system-wide work, large migrations, repo-wide audits, and high-confidence verification prompts to /hyperflow:workflow. In Claude Code v2.1.154+, it asks the native dynamic workflow runtime to create a background workflow with research, parallel execution, adversarial verification, quality gates, and final synthesis. In Codex and OpenCode, it runs the same phases through a portable workflow adapter using provider subagents/tasks when available and inline worker/reviewer phases otherwise.
A growing collection of Claude-compatible academic workflow bundles. Covers scientific figures, manuscript writing and polishing, reviewer assessment, citation retrieval, data availability, paper reading, literature search, response letters, paper-to-PPTX conversion, and evidence-grounded Chinese invention patent drafting. Rules are organized as reusable skill folders with explicit workflows and quality checks.
Harness-native ECC plugin for engineering teams - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, MCP conventions, and operator workflows for Claude Code plus adjacent agent harnesses
Develop, test, build, and deploy Godot 4.x games with Claude Code. Includes GdUnit4 testing, web/desktop exports, CI/CD pipelines, and deployment to Vercel/GitHub Pages/itch.io.
'%20stop-opacity%3D'0.16'%2F%3E%3Cstop%20offset%3D'1'%20stop-color%3D'rgb(200%2C90%2C60)'%20stop-opacity%3D'0.03'%2F%3E%3C%2FlinearGradient%3E%3C%2Fdefs%3E%3Crect%20width%3D'320'%20height%3D'200'%20fill%3D'url(%23g)'%2F%3E%3Ccircle%20cx%3D'250'%20cy%3D'56'%20r%3D'92'%20fill%3D'rgb(200%2C90%2C60)'%20fill-opacity%3D'0.06'%2F%3E%3Ccircle%20cx%3D'64'%20cy%3D'172'%20r%3D'58'%20fill%3D'rgb(200%2C90%2C60)'%20fill-opacity%3D'0.05'%2F%3E%3C%2Fsvg%3E)
