plugin-introspector

White-box monitoring of Claude Code plugin workflows exposes tool usage, token consumption, and execution patterns. It identifies bottlenecks, waste, and security risks to provide actionable optimization opportunities across multi-agent pipelines.

Data-driven improvement proposals are generated by an orchestrated pipeline that analyzes plugin execution. It enforces meta-rules validation and applies ROI-guided improvements to optimize performance, security, and reliability automatically.

Yes, the introspector plugin evaluates compliance and security across multi-agent pipelines. It provides end-to-end visibility into execution patterns and enforces meta-rules to detect security risks and ensure workflow compliance.

Yes, jq is a required dependency to run the introspector plugin scripts. The scripts rely on jq to process and parse the JSON data extracted from Claude Code plugin workflows during monitoring and analysis.

White-box monitoring identifies optimization opportunities by analyzing tool usage, prompts, and token consumption in real-time. It exposes hidden inefficiencies within the execution patterns and generates proposals guided by ROI metrics and meta-rules.

Meta-rules validation is limited to the predefined constraints configured within the introspector's analysis pipeline. Optimization proposals are strictly ROI-guided, meaning changes without a clear performance, security, or reliability return are filtered out.