help

The help skill serves as the central command and discovery interface for the Continuous-Claude environment. It is designed to minimize cognitive load by allowing users to express their intent in natural language while the system maps that intent to the most effective combination of agents, workflows, and tools. Whether a user is performing codebase exploration, debugging, feature implementation, or formal mathematical verification, this skill guides them to the right resources.

• Provides interactive discovery through targeted user questions to identify specific development goals.

• Catalogues specialized agents including architectural planners like architect and phoenix, implementation agents like kraken, and investigation agents like sleuth and debug-agent.

• Documents workflow orchestration for multi-agent pipelines such as /fix for bug remediation and /build for feature development.

• Offers detailed references for CLI tools including TLDR code analysis for efficient token-usage, formal verification via /prove and Godel-Prover, and memory management via recall systems.

• Explains active lifecycle hooks that automate session registration, context preservation, and task-specific suggestions during the UserPromptSubmit and SessionStart phases.

• Supports deep-dive documentation for specific components, allowing users to query individual skill or agent capabilities.

• Invoke with /help for an interactive interview or provide an argument like /help workflows or /help agents for specific references.

• Use the goal-oriented options (e.g., Fix a bug, Build a feature) to trigger optimized tool suggestions based on the current codebase state.

• Integrates with memory systems to recall past learnings and session-specific hooks to ensure compliance with the repository's rules and frontmatter configurations.

• Note that some capabilities, such as /prove, require external infrastructure like LM Studio, while tools like tldr are natively supported for token-efficient codebase traversal.