Claude Code skills ecosystem for a team
Context
Moving from individual use of Claude Code to an agentic workflow shared across a whole team, reproducible: that's the starting point, once the tool is adopted beyond a single seat.
More than 30 skills, built for a team rather than a single seat, cover document production, research, orchestration and MCP servers.
In the background, an onboarding constraint: an install script, an onboarding README and a CLAUDE.md template per engagement.
The point is not to add one more skill: it's to reason about execution boundaries so the tooling holds at team scale, not just for a single seat.
The system
An explicit boundary organizes the whole thing: what runs inside the agent on one side, what stays usable on the conversational-assistant side on the other.
Equip a team, not just yourself: the principle holds for the whole ecosystem, from the skills themselves to how a new engagement inherits them.
Decisions & trade-offs
More than 30 skills, automatic install, an onboarding README, a CLAUDE.md template per engagement.
More than 30 Claude Code skills for a team: document production, research, orchestration, MCP servers. With an install script, an onboarding README and a CLAUDE.md template per engagement.
Separate what runs inside the agent, skills, hooks, MCP, from what stays portable on the assistant side.
The architecture separates what runs inside the agent, skills, hooks, MCP, from what stays usable on the conversational-assistant side.
Result
A platform for a team, not a throwaway script.
What it shows
Equipping a team on an agentic workflow and reasoning about execution boundaries.