parallel-agents

Parallel-agents is a robust orchestration framework within the Antigravity Kit designed to manage complex software development lifecycles by invoking domain-specific AI personas. Instead of relying on a single general-purpose AI, this skill allows users to trigger a team of specialized agents—such as security auditors, backend specialists, frontend engineers, and devops experts—to collaborate on a single task. This pattern is ideal for comprehensive code audits, multi-layered feature implementations, and system architecture reviews where diverse expertise is required for high-quality output. The skill supports sequential chains, context-aware handoffs, and final synthesis, ensuring that findings from multiple agents are aggregated into a single, cohesive report.

• Coordinates 17+ specialized agents including security, performance, database, and documentation experts.

• Facilitates complex multi-step workflows like comprehensive codebase mapping, vulnerability penetration testing, and automated test generation.

• Manages state and context persistence, allowing findings from a frontend specialist to inform subsequent backend or test engineer actions.

• Implements a standardized Synthesis Protocol to generate unified, actionable summaries, including critical recommendations and prioritized task lists.

• Enables native agent invocation through simple natural language commands, making it accessible for rapid prototyping and enterprise-grade refactoring.

• Best suited for complex architectural tasks, large-scale refactors, and end-to-end feature implementations; avoid using for trivial, single-domain code fixes.

• Always utilize the synthesis phase to consolidate agent outputs into a single report to maintain clarity and avoid fragmented information.

• Ensure logical ordering in chains, typically starting with exploration or discovery agents before moving to domain-specific implementation or analysis.

• Combine with built-in Antigravity agents like Explore and Plan to leverage high-speed codebase indexing alongside custom domain expertise.

• Maintain high-quality context passing by explicitly referencing previous agent outputs when initiating subsequent tasks to maximize AI efficacy.