Compositions

Settings + Project Instructions

Settings enforce available capabilities; instructions explain intended use and remediation. Generate both from one operating policy where practical.

Settings + Sandbox

Use product permissions for usable workflow and OS, container, or hosted isolation for containment. Test the boundary rather than trusting a mode name.

Settings + Hooks

Settings activate and constrain hooks; hooks provide lifecycle checks. Keep a protected external gate for consequential outcomes.

Settings + MCP Policy

Client settings decide which servers and tools are visible. Server-side identity and authorization decide what those tools can actually do.

Settings + Subagents

Define whether child agents inherit, narrow, or override parent tools and approvals. Least privilege should become stricter with specialization, not broader.

Settings + Release Authorization

The agentAgentA worker you delegate to: brief it, and it takes steps on its own. A chatbot answers; an agent acts.An LLM that runs tools in a loop toward a goal — it acts, checks the result, decides the next step, and repeats until done.Full definitionClose may prepare and verify a change. A deterministic policy layer evaluates admissible evidence and a named human authorizes release when required.