observer

To maintain durable project memory without altering product code, you can track notes, drift, and risks in repository artifacts. This approach stores observations and metrics in JSONL files and generates syntheses in markdown, providing a persistent project-health view across sessions.

Project drift detection tracks long-term health by recording passive, additive observations of patterns and decisions over time. It surfaces these insights by storing observations in data files and synthesizing them into a durable project-intelligence markdown document.

You record observations and metrics for risk tracking by appending entries to data/observations.jsonl and data/metrics.jsonl. This passive, additive approach ensures evidence-backed reporting by storing simple metrics and contextual notes directly in repository artifacts.

Yes, project memory tracking works without modifying source code by storing all data in repository-owned artifacts. It records context in JSONL files and generates syntheses in a dedicated docs directory, ensuring the product code remains completely unaltered.

You should use repository artifacts for session memory when you need to recall context across multi-session workflows without modifying product code. This approach is ideal for tracking drift and risks passively while maintaining an evidence-backed project-health view.

The best way to synthesize project intelligence from tracked notes and drift is to aggregate observations and metrics into a single markdown file. This generates a durable summary in docs/observer/project-intelligence.md, providing a comprehensive project-health view with guardrails for evidence-backed reporting.