Does this skill prevent the AI from making its own decisions?

Yes, it explicitly forbids subagents from making assumptions, inventing requirements, or deciding on next steps beyond their immediate, assigned task.

How does this skill improve multi-agent workflows?

It eliminates ambiguity by requiring agents to signal DONE when finished or BLOCKED when they need more information, preventing silent failures and scope creep.

Is this skill meant to be triggered by the user?

No, it is a behavioral constraint skill designed to be loaded into subagents to govern their internal logic and interaction with a supervisor agent.

What is the purpose of the Subagent Contract in Claude Code?

It serves as a behavioral framework that forces specialized AI agents to stay within their assigned scope and use specific status signals for reliable orchestration.

When should a subagent return a BLOCKED status?

An agent should signal BLOCKED immediately if information is missing, requirements are unclear, or it lacks the necessary inputs to proceed without guessing.

Subagent Contract

Name: Subagent Contract
Author: Jamie-BitFlight

byJamie-BitFlight

•

Productivity & Workflow

Enforces strict role boundaries and status signaling for specialist AI agents within multi-agent workflows.

About

The Subagent Contract skill establishes a standardized framework for specialist agents to operate reliably within a multi-agent orchestration system. It prevents scope creep and unauthorized decision-making by enforcing rigid identity constraints and mandatory status signaling protocols. By implementing standardized DONE and BLOCKED signals, this skill ensures that subagents follow Standard Operating Procedures (SOPs) exactly, report deliverables predictably, and flag missing information immediately rather than making risky assumptions, leading to more stable and manageable autonomous agentic workflows.

Key Features

Mandated STATUS: BLOCKED signaling to flag missing inputs or ambiguities immediately
Quality verification checklists for both task completion and blocking scenarios
Mandatory STATUS: DONE signaling with delivery summaries and risk identification
14 GitHub stars
Strict identity constraints to prevent subagents from expanding scope or inventing requirements
Standardized supervisor interaction protocol for task acknowledgment and results reporting

Use Cases

Ensuring predictable behavior and structured output formats in automated agentic workflows
Preventing unauthorized architectural changes or 'hallucinated' features during autonomous runs
Orchestrating complex software projects where tasks are delegated to multiple specialized subagents

About

Key Features

Mandated STATUS: BLOCKED signaling to flag missing inputs or ambiguities immediately
Quality verification checklists for both task completion and blocking scenarios
Mandatory STATUS: DONE signaling with delivery summaries and risk identification
14 GitHub stars
Strict identity constraints to prevent subagents from expanding scope or inventing requirements
Standardized supervisor interaction protocol for task acknowledgment and results reporting

Use Cases

Ensuring predictable behavior and structured output formats in automated agentic workflows
Preventing unauthorized architectural changes or 'hallucinated' features during autonomous runs
Orchestrating complex software projects where tasks are delegated to multiple specialized subagents