ask

The ask skill functions as a forensic interface for your codebase, allowing you to bridge the gap between static code and the dynamic thought process that created it. By leveraging git-ai metadata, this skill identifies the specific AI model, session, and prompt sequence responsible for generating selected blocks of code. Instead of relying on manual investigation or guessing why a complex function was structured in a specific way, you can engage with the implicit authorial intent recorded during the development lifecycle. This is particularly valuable for complex refactoring, onboarding into unfamiliar modules, or verifying architectural patterns in large repositories. When you invoke this skill, it performs a context-aware lookup, cross-referencing your editor selection with recorded git-ai history to reconstruct the context of the initial request.

• Automatically maps code segments to original prompts and conversational transcripts using git-ai.

• Employs a specialized subagent to interpret technical decisions, such as recursion vs. iteration or specific library selections.

• Provides first-person explanations from the 'author agent' perspective to offer deeper insight into the problem-solving process.

• Seamlessly integrates with editor selection context, allowing you to ask questions directly from your workspace without manual file navigation.

• Offers fallback analytical modes when historical AI data is missing, ensuring you always receive an objective code review if a direct prompt match cannot be established.

• To use this skill, select the specific lines or function in your IDE and trigger the /ask command with your natural language query.

• The skill expects a specific path and line range; if none are provided, it will prompt you to clarify the target selection.

• It is designed for fast, local-first retrieval, utilizing bash-based git-ai commands to ensure your data stays private and secure.

• Avoid manually searching through session logs or JSONL files; let the subagent handle the parsing and synthesis of the conversation data.

• If the tool finds no AI history, it will revert to objective code analysis, providing a technical breakdown of the logic rather than a simulated author perspective.