soul

You can model user work patterns by collecting session signals with context and raw evidence, then synthesizing inferred patterns into an operational profile that grounds downstream actions in observed behavior rather than generic assumptions.

You can apply context-aware communication preferences by using the propose command in hybrid mode to generate a SOUL.md diff with evidence citations, reviewing the changes, and applying them only after passing strict drift detection and approval guardrails.

Drift detection works by blocking large or contradictory changes to the user profile unless explicitly acknowledged, resolving contradictions as conditional rules to maintain an accurate, evidence-backed identity synthesis without sudden unreviewed shifts.

Yes, you can generate shipping cadence context by running retro commands that ground summaries in retro metrics and synthesized work-pattern models, producing an operational profile that reflects actual observed shipping behaviors.

Yes, you need YAML frontmatter-defined commands to initialize, learn, propose, apply, diff, and reset the session memory profile, as these commands enforce the strict thresholds, anti-generic filtering, and drift detection required for evidence-backed synthesis.

Your user profiling synthesis gets blocked by the anti-generic filter when proposed updates lack sufficient raw evidence or fall below minimum thresholds, preventing generic assumptions from contaminating the evidence-backed identity profile.