Commands
Agile Work
cd my-project
agileflow init
# Add individual workflows
agileflow add diagnosing-bugs
agileflow add filing-pr
# Keep them current
agileflow updateThen open Codex, Claude Code, Cursor, OpenCode, or Gemini CLI as usual. When a request matches a skill's description (for example, "the profile page crashes when I upload a big image"), the provider loads the skill by itself. Manual skills load only when you invoke them by name.
What init creates
repo/
├── agileflow.yaml
├── agileflow.lock
├── .agents/
│ └── skills/
│ ├── diagnosing-bugs/
│ ├── checking-blast-radius/
│ └── verifying-changes/
└── .claude/
└── skills/ # only if Claude Code is detected
├── diagnosing-bugs -> ../../.agents/skills/diagnosing-bugs
└── ...That is all. AgileFlow does not create a docs/ hierarchy, does not change AGENTS.md, CLAUDE.md, or GEMINI.md, installs no hooks, and changes no provider settings.
Share with your team
Commit agileflow.yaml, agileflow.lock, and .agents/skills/. Teammates get the skills with the repository. After git pull or switching branches, agileflow sync makes their files match the lockfile exactly.
