Your AI agent's project manager is called `vibe`

What it does

VibeSkills is a Python-based orchestration layer—self-described as a “Super Skill”—that wraps around AI agents to impose a working rhythm. After you invoke the vibe entry point, the harness freezes your intent, stages a plan, routes tasks to appropriate expert Skills from a library of 340-plus, verifies the results with tests or artifacts, and preserves workspace memory for the next session. The goal is to replace endless chat drift with a structured path from idea to delivery.

The interesting bit

The project treats workflow discipline itself as a skill. Instead of adding more isolated tools, it defines how work ought to run and invites future domain-specific Skills to plug into the same pipeline. It is essentially a meta-layer of glue code with ambition: it borrows structure from projects like Superpowers and GSD, but packages them as a portable, extensible runtime.

Key highlights

• Single entry point (vibe) governs the full lifecycle from intent to verified delivery.
• Automatic routing across 340-plus bundled Skills bounded by phase—requirements, planning, implementation, review, and cleanup.
• Workspace memory persists requirements, plans, and evidence across sessions.
• Enforces a “proof trail”: work must pass tests, checks, or explicit manual review before the harness marks it complete.
• Designed as an “open skill plane,” so future domains like research, finance, or law can theoretically join the same harness.

Caveats

• The README is rich with diagrams and glossary, but sparse on implementation specifics—the exact mechanics of the VCO runtime and how the router selects a Skill remain unclear.
• The 340-plus Skill count appears only as a badge; no manifest or catalog is visible in the provided text.
• Because it is fundamentally a workflow wrapper, its value hinges on whether your current agent stack lacks staging, memory, and verification discipline.

Verdict

Try it if your AI agent lacks staging, memory, or verification discipline. Pass if you already have a rigid execution loop and prefer to hand-manage every Skill call.