💬 Join the Discord — discussions, updates, feature requests, bug reports
Flow-Next is an AI agent orchestration plugin. Sixteen agent-native skills for the full lifecycle: idea → spec → tasks → review → ship → maintain. Bundled task tracking, dependency graphs, re-anchoring, multi-model reviews, decay-aware project memory, GitHub PR resolution, agent-readiness audits. Everything lives in your repo — no external services, no global config. Uninstall: delete .flow/.
First-class on Claude Code, OpenAI Codex (CLI + Desktop), and Factory Droid. Also runs on OpenCode via the community port.
🆕 v0.38.0 — Capture + interview grill-me. New
/flow-next:capturesynthesizes free-form discussion into a flow-next epic spec with source-tagged criteria + mandatory read-back./flow-next:interviewnow leads with recommendations + confidence tiers, investigates the codebase before asking, walks decision trees in dependency order. Cross-platform tool handling moved into the Codex sync script — canonical skills stay Claude-native, sync rewrites for Codex mirror. Full changelog.
🌐 Visual overview at mickel.tech/apps/flow-next — diagrams, examples, the full feature tour.
| Claude Code | OpenAI Codex | Factory Droid |
/plugin marketplace add \
https://github.com/gmickel/flow-next
/plugin install flow-next
/flow-next:setup |
git clone https://github.com/gmickel/flow-next.git
cd flow-next
./scripts/install-codex.sh flow-next
# then: /flow-next:setup |
droid plugin marketplace add \
https://github.com/gmickel/flow-next
# /plugins → install flow-next |
Why a script for Codex? Codex's plugin protocol currently only registers skills from plugin.json — not custom .toml agents or hooks. The /plugins install gives you slash commands, but no subagent isolation (worker model tier, disallowed_tools) and no Ralph hooks. install-codex.sh merges all 21 agents + hooks directly into ~/.codex/config.toml so you get the full multi-agent + Ralph experience. We'll switch to /plugins once Codex's manifest supports agents and hooks fields.
Update Codex: cd flow-next && git pull && ./scripts/install-codex.sh flow-next. The script is idempotent — safe to re-run on every update.
📖 Full docs · Codex install guide · OpenCode port
flowchart LR
Idea([💡 Idea]) --> P[/flow-next:prospect/]
Idea --> C[/flow-next:capture/]
P --> C
P -.->|direct via promote| L[/flow-next:plan/]
C --> L
C --> I[/flow-next:interview/]
I --> L
L --> W[/flow-next:work/]
W --> R[/flow-next:impl-review/]
R -->|SHIP| Done([🚀 Ship])
R -->|NEEDS_WORK| W
Done -.maintenance.-> A[/flow-next:audit/]
A -.-> M[(.flow/memory/)]
Idea → spec → tasks → ship. Branch in, branch out — pick the entry point that matches your context.
| Problem | Solution |
|---|---|
| Context drift | Re-anchoring before every task — re-reads specs + git state |
| Context window limits | Fresh context per task — worker subagent starts clean |
| Single-model blind spots | Cross-model reviews — RepoPrompt, Codex, or Copilot as second opinion |
| Forgotten requirements | Dependency graphs — tasks declare blockers, run in order |
| "It worked on my machine" | Evidence recording — commits, tests, PRs tracked per task |
| Infinite retry loops | Auto-block stuck tasks — fails after N attempts, moves on |
| Duplicate implementations | Pre-implementation search — worker checks for similar code before writing new |
| Hallucinated specs from "I think we discussed…" | Source-tagged capture — every acceptance criterion marked [user] / [paraphrase] / [inferred], mandatory read-back loop |
| Stale project memory polluting future work | /flow-next:audit + categorized memory schema — agent reviews each entry, flags stale (never deletes) |
| GitHub PR review threads piling up | /flow-next:resolve-pr — fetch → triage → dispatch resolver agents → reply → resolve via GraphQL |
| Command | What It Does |
|---|---|
/flow-next:prospect |
Generate ranked candidate ideas grounded in the repo, upstream of capture/interview/plan |
/flow-next:capture |
Synthesize conversation context into an epic spec (source-tagged, mandatory read-back) |
/flow-next:interview |
Deep spec refinement with lead-with-recommendation + confidence tiers + codebase-first investigation |
/flow-next:plan |
Research codebase, create epic + dependency-ordered tasks |
/flow-next:work |
Execute tasks with re-anchoring + worker subagents + review gates |
/flow-next:impl-review |
Cross-model implementation review (RepoPrompt, Codex, or Copilot) |
/flow-next:plan-review |
Cross-model plan review |
/flow-next:epic-review |
Epic-completion review gate — verify combined implementation matches spec |
/flow-next:resolve-pr |
Resolve GitHub PR review threads (fetch → triage → fix → reply → resolve via GraphQL) |
/flow-next:audit |
Agent-native review of .flow/memory/ entries against current code (Keep / Update / Consolidate / Replace / Delete) |
/flow-next:memory-migrate |
Lift legacy flat memory files into the categorized schema; agent classifies each entry |
/flow-next:prime |
8-pillar agent-readiness assessment with parallel scouts; remediation via consent prompts |
/flow-next:ralph-init |
Scaffold autonomous loop (scripts/ralph/) |
Run overnight. Fresh context per iteration + multi-model review gates + auto-block stuck tasks.
/flow-next:ralph-init # One-time setup
scripts/ralph/ralph.sh # Run from terminal📖 Ralph deep dive · Ralph TUI (bun add -g @gmickel/flow-next-tui)
| Project | Platform |
|---|---|
| flow-next-opencode | OpenCode |
| FlowFactory | Factory.ai Droid |
GNO — Local hybrid search for your notes, docs, and code. Give Claude Code long-term memory over your files via MCP.
bun install -g @gmickel/gno && gno mcp install --target claude-code
Made by Gordon Mickel · @gmickel · gordon@mickel.tech
