sergvss

setup-agent-team

Assemble an AI agent team in the current project - pick the roles, copy the templates into .claude/agents/, adapt them to the project stack and verify the result. Use when asked to set up an agent team, add a role to an existing team, or deploy the agents-best-teams methodology in a project.

context-checklist

Context management in an agent team - what to pass into a subtask, what to return from it, what to write to memory and what never to write. Use when handing work to another agent, when preparing a result report, and when the context is approaching its limit.

implementation-path

The path from task to commit - the sequence of implementation steps with a check at each one. Use when carrying out a task of Tier 2 or above, where the work must reach a verified and committed result rather than just written code.

permission-checklist

Permission matrix for an agent team - which tools, permissionMode, maxTurns, effort and memory each role is entitled to. Use when adding a new role, when reviewing an existing one, and after an incident where an agent stepped outside its area.

planning-checklist

Task decomposition before work starts - determine the Tier, split into subtasks, assign roles, set done criteria. Use when a task is larger than a single edit, when it is unclear where to begin, and always before delegating work to other agents.

tool-checklist

Check before calling a tool - is it needed at all, is it the right one, how large is the output, what is the risk class. Use before Bash with destructive potential, before bulk edits, before database queries, and whenever a tool's output could turn out to be enormous.