FlowState Skills is a complete agent-driven methodology for working inside the FlowState platform — entity management, registration, multi-phase planning, task execution, code review, business plan and product creation, agent inbox triage, and process automation — packaged as a set of composable skills with the bootstrap instructions that make sure your agent uses them.

Give your agent FlowState Skills: Claude Code · Codex CLI · Cursor · Gemini CLI · OpenCode · GitHub Copilot CLI · Factory Droid.

From the moment the session starts, the FlowState bootstrap skill is injected into your agent's context. The agent sees the full FlowState skill catalog and the absolute rules — never fabricate IDs, always read .flowstate/config.json before any MCP call, always resolve agent identity before dispatching subagents.

When you ask the agent to onboard a repo, build a feature, triage its inbox, or spin up a product and business plan, it reaches for the matching skill instead of guessing. Skills compose: brainstorming hands off to multi-phase planning, which hands off to task execution, which hands off to code review and branch finishing — each step gated by the right approval, identity, and time-tracking sub-skills.

The agent checks for relevant skills before any task. They are mandatory workflows, not suggestions.

Installation differs by harness. If you use more than one, install FlowState Skills separately for each.

FlowState Skills is distributed via the Epic Digital marketplace.

• Register the marketplace:

  /plugin marketplace add epic-digital-im/epic-flowstate-skills

• Install the plugin:

  /plugin install flowstate-skills@epic-flowstate

• Open the plugin search interface:

  /plugins

• Search for flowstate-skills and select Install Plugin.

If FlowState Skills is not yet available in the official Codex marketplace, install directly from this repository by cloning it and pointing your Codex config at .codex-plugin/plugin.json.

For local development, keep Codex's installed skill copies synchronized from this repository:

node scripts/sync-codex-skills.mjs --dry-run
node scripts/sync-codex-skills.mjs

The repository's skills/ directory is the source of truth; ~/.codex/skills is a generated local install target.

• In Cursor Agent chat, install from the marketplace:

  /add-plugin flowstate-skills
  

• Or search for "flowstate-skills" in the plugin marketplace.

If FlowState Skills is not yet listed, clone this repository and point Cursor at .cursor-plugin/plugin.json.

• Install the extension:

  gemini extensions install https://github.com/epic-digital-im/epic-flowstate-skills

• Update later:

  gemini extensions update flowstate-skills

OpenCode uses its own plugin install; install FlowState Skills separately even if you already use it in another harness.

• Tell OpenCode:

  Fetch and follow instructions from https://raw.githubusercontent.com/epic-digital-im/epic-flowstate-skills/main/.opencode/INSTALL.md
  

• Detailed docs: .opencode/INSTALL.md

• Register the marketplace:

  copilot plugin marketplace add epic-digital-im/epic-flowstate-skills

• Install the plugin:

  copilot plugin install flowstate-skills@epic-flowstate

• Register the marketplace:

  droid plugin marketplace add https://github.com/epic-digital-im/epic-flowstate-skills

• Install the plugin:

  droid plugin install flowstate-skills@epic-flowstate

1. flowstate-using-flowstate-skills — Loaded at session start. Establishes the absolute rules (never fabricate IDs, always read config) and the full skill catalog so every other skill is reachable by name.

2. flowstate-workspace-registration / flowstate-codebase-registration / flowstate-monorepo-audit — Activates when a repo is not yet linked to FlowState. Creates workspace, codebase, and per-package project records and writes .flowstate/config.json files.

3. flowstate-brainstorming — Activates before writing code. Refines rough ideas through Socratic discussion, presents the design in chunks for human approval, and saves an approved design spec document.

4. flowstate-multi-phase-planning — Activates with an approved design. Decomposes the spec into one milestone with N phase tasks, each scoping goals, deliverables, and acceptance criteria.

5. flowstate-executing-multi-phase-plan — Iterates phase tasks in sortOrder and runs the full task execution lifecycle for each.

6. flowstate-task-execution — The 10-step lifecycle: pre-flight check, worktree isolation, plan, subagent-driven development, two-stage review, PR creation, merge, time tracking, verification.

7. flowstate-test-driven-development — Enforced during implementation. RED-GREEN-REFACTOR with a failing test before production code.

8. flowstate-code-review / flowstate-receiving-code-review — Per-task and per-milestone review against the design spec. Critical issues block progress.

9. flowstate-finishing-milestone / flowstate-completing-milestone — Milestone-level review against spec, followup tasks, and final completion.

10. flowstate-finishing-branch — Verifies tests, presents merge/PR/keep/ discard options, cleans up the worktree.

The agent checks for relevant skills before any task. Mandatory workflows, not suggestions.

• flowstate-org-schema, flowstate-workspace-schema, flowstate-codebase-schema, flowstate-project-schema, flowstate-milestone-schema, flowstate-task-schema — Field, ID, and relationship reference for every collection.
• flowstate-object-hierarchy, flowstate-config-validation — How IDs inherit through .flowstate/config.json files; how to diagnose missing IDs.

• flowstate-workspace-registration, flowstate-codebase-registration, flowstate-monorepo-audit, flowstate-create-if-missing, flowstate-pre-flight-check.

• flowstate-brainstorming, flowstate-multi-phase-planning, flowstate-executing-multi-phase-plan, flowstate-task-execution, flowstate-finishing-milestone, flowstate-completing-milestone, flowstate-project-audit, flowstate-agent-inbox, flowstate-agent-collaboration-routing, flowstate-agent-handoff-response, flowstate-agent-conversation-protocol, flowstate-document-creation, flowstate-visual-companion, flowstate-agent-runner-operations, flowstate-skill-to-process-conversion.

• flowstate-writing-plans, flowstate-subagent-development, flowstate-code-review, flowstate-receiving-code-review, flowstate-finishing-branch, flowstate-test-driven-development, flowstate-systematic-debugging, flowstate-dispatching-parallel-agents, flowstate-using-git-worktrees, flowstate-verification-before-completion.

• flowstate-product-schema, flowstate-product-create, flowstate-product-link-projects, flowstate-product-link-teammembers, flowstate-bizplan-schema, flowstate-bizplan-create, flowstate-bizplan-link-product, flowstate-bizplan-link-teammembers, flowstate-creating-a-product, flowstate-creating-a-bizplan.

• flowstate-agent-identity, flowstate-entity-assignment, flowstate-agent-collaborator-lookup, flowstate-agent-task-assignment, flowstate-agent-mentioning, flowstate-entity-audit, flowstate-timetracking, flowstate-approval-workflow.

• flowstate-agentmemory-tools, flowstate-agentmemory-workflow — Harness-neutral AgentMemory tool reference and recall-to-verify workflow for shared memory, agent-scoped continuity, and safe durable writes.

• flowstate-cli-local-auth, flowstate-cli-cloud-auth, flowstate-cli-wallet-auth, flowstate-saga-wallet, flowstate-cloud-gateway-routing, flowstate-agent-cli-bootstrap, flowstate-dojo-cli, flowstate-plugin-lifecycle, flowstate-saga-skill-record, flowstate-cloud-deployment.

• flowstate-process-registration, flowstate-process-execution.

• flowstate-writing-skills, flowstate-using-flowstate-skills, flowstate-subtask-template.

For the full catalog with triggers and dependency map, see skills/flowstate-using-flowstate-skills/SKILL.md.

• Never fabricate IDs — every FlowState ID comes from config, an MCP response, or the user. No exceptions.
• Read config before every MCP call — orgId, workspaceId, codebaseId, projectId always come from .flowstate/config.json.
• Resolve agent identity before dispatch — every subagent and every entity creation is attributed to a teamMemberId.
• Test-Driven Development — write tests first, always.
• Systematic over ad-hoc — process over guessing.
• Evidence over claims — verify before declaring success.

.
├── .claude-plugin/        Claude Code plugin manifest + marketplace
├── .codex-plugin/         Codex CLI plugin manifest
├── .cursor-plugin/        Cursor plugin manifest
├── .opencode/             OpenCode plugin (JS) + INSTALL.md
├── gemini-extension.json  Gemini CLI extension manifest
├── GEMINI.md              Gemini context file (loads bootstrap skill)
├── hooks/                 Cross-platform SessionStart hook
├── skills/                The full FlowState skill library
├── package.json           npm metadata for OpenCode git+https install
├── AGENTS.md              Symlink to CLAUDE.md (agents.md standard)
├── CLAUDE.md              Contributor + agent guide
└── README.md

FlowState Skills updates are mostly harness-dependent. Most harnesses pick up new versions automatically; OpenCode and some npm-pinned setups may need a plugin reinstall. See each harness's install section above.

1. Fork the repository.
2. Create a branch from main.
3. Follow the flowstate-writing-skills skill for any new or modified skill.
4. Test on at least one harness — ideally Claude Code and one other.
5. Bump the version in lockstep across every plugin manifest (see CLAUDE.md).
6. Submit a PR describing the problem you solved and the harnesses you tested.

We do not generally accept project-specific or fork-specific skills. Domain skills belong in their own plugin.

MIT License — see LICENSE.

• Issues: https://github.com/epic-digital-im/epic-flowstate-skills/issues
• Maintainer: Spencer Thornock / Epic Digital Media