session-reflection

Analyzing session history identifies recurring prompt gaps and workflow blockers, such as repeated corrections or restarts. This reflection process categorizes interventions to help teams turn recurring issues into durable system prompt improvements and crash-resilient workflows.

You should analyze session history and refine system prompts after hitting milestones, experiencing repeated corrections, or following restarts. These events indicate workflow blockers or context drift, signaling the need to update the persistent working state and prompt guidelines.

To generate a persistent working state, analyze session history including conversations and memory files to categorize interventions. This produces a WORKING_STATE.md file that captures the current progress and integrates with a launcher pattern to maintain governance across long-running sessions.

Yes, applying a memory protocol categorizes session interventions into a structured taxonomy. This categorization identifies recurring gaps between user expectations and agent behavior, generating pattern-based improvements to system prompts for better long-running session resilience.

Session reflections depend on the availability and completeness of session history data like conversations and git logs. If memory files lack sufficient detail or interventions are not properly categorized, the generated working state and system prompts may fail to prevent context drift effectively.