Correction Tracking
Monitors and logs user corrections to refine AI workflows and prevent repetitive mistakes over time.
The Correction Tracking skill is a foundational component of the Forge framework designed to capture 'correction moments' where users redirect, correct, or refine agent output. By systematically logging these instances—such as template gaps, context misunderstandings, or assumption errors—in a structured retrospective format, it enables continuous improvement of AI-driven engineering processes. This skill ensures that agents learn from their mistakes in-session and provide data-driven insights for long-term prompt and workflow optimization via the Forge improvement cycle.
Key Features
01Structured logging to retrospective-notes.md with metadata
02Status tracking for incremental processing and audit review
03Seamless integration with the /forge:improve optimization command
041 GitHub stars
05Real-time detection of user redirection and feedback signals
06Automated root cause analysis for agent misunderstandings
Use Cases
01Providing a transparent audit trail of agent-user interactions for team retrospectives
02Identifying recurring template gaps to improve initial project scaffolding
03Analyzing and fixing context-aware errors in complex system architectures