Superpowers SSoT

Superpowers (Procedural Agentic Skills)

Superpowers are high-level, structured procedural workflows that enforce disciplined methodologies across the Vox agentic network. Unlike basic tool calls, Superpowers are multi-phase execution graphs that orchestrate multiple agents and tools to achieve complex engineering goals.

Core Philosophy

1. Planning-First: No Superpower may execute code without a verified and approved implementation plan.
2. Verification-Locked: Completion is only granted when automated tests (TDD) and architectural audits pass.
3. Cross-Agent Orchestration: Skills like Research and Review can trigger handoffs between specialized agents (e.g., Lane G for research synthesis).

The 14 Standard Superpowers

Skill	Category	Description
Brainstorm	Strategic	High-level ideation and problem space exploration.
Specify	Strategic	Formalization of requirements into XML/Markdown specs.
Plan	Strategic	Architecture of the execution graph and task decomposition.
TDD	Execution	Red-Green-Refactor loop implementation.
Debug	Execution	Root-cause analysis and automated repair cycle.
Refactor	Execution	Improving structure without behavior change.
Review	Verification	alignment check against specifications and lint rules.
Research	Strategic	Autonomous context gathering from web/local corpora.
Mockup	Execution	Browser-based UI prototype generation.
Audit	Verification	Security and architectural boundary validation.
Document	Execution	Synchronization of implementation and documentation.
Optimize	Execution	Performance profiling and hotspot tuning.
Sync	Strategic	VCS management and conflict resolution.
Deploy	Execution	build, test, and release orchestration.

Integration Wiring

• SkillRegistry: Tracks installed Superpowers and their associated prompts/instructions.
• Orchestrator: Injects Superpower-specific continuation instructions into the agent context.
• GUI Discovery: The SuperpowersCatalog provides a visual interface for activating and configuring these workflows.

Safety & Guardrails

• Budgeting: Superpowers are subject to financial and attention budgets.
• Rollout: Experimental Superpowers use a rollout percentage (default 0%) until hardened.
• Shadow Mode: Superpowers can run in observation mode to collect telemetry without mutating the repository.