flowai-reflect-by-history

Session log analysis detects recurring agent behavior patterns by parsing historical IDE transcripts to extract tool usage, errors, and timestamps. It identifies repeated manual workflows and missing primitives, producing a classified report with evidence and recommended fixes.

The best way to detect repeated manual workflows is cross-session pattern detection applied to historical OpenCode transcripts. It extracts primitive invocations and tool usage sequences, classifies recurring ad-hoc decisions, and maps corrective actions to Rules, Skills, or Commands.

Yes, you can review project-local Claude session files to identify missing documentation. The analysis parses transcripts to extract user requests and primitive invocations, detects unchecked invariants and documentation gaps, and maps corrective actions to Project Docs or Code Changes.

Generate a prioritized report by reading JSONL session transcripts to extract tool usage, errors, and timestamps. Cross-session pattern detection classifies recurring issues with frequency counts and evidence, recommending specific artifacts like Hooks or Rules to implement durable fixes.

No, historical session analysis focuses exclusively on previous sessions and explicitly excludes the active live session. It reads local JSONL or SQLite session files to surface recurring errors and systemic workflow issues without fabricating findings when history is absent.

Analyzing agent behavior patterns across past sessions surfaces systemic workflow issues and gaps in project primitives, allowing teams to prioritize durable fixes. It classifies recurring errors and ad-hoc decisions, mapping them to specific artifacts like Rules, Skills, or Commands instead of repeated remedies.