Project Memory: Cross-Session Context Retention

A behavioral skill for maintaining a lightweight, per-project memory so personas don't re-derive already-known facts, decisions, and corrections across sessions. This is not a backup mechanism — it's a practice of writing and reading context during development work.

What Is This?

As you work on a project across multiple sessions, you and the user may:

• Discover project-specific patterns or conventions worth remembering (e.g., "always run migrations with the wrapper script, not raw sql")
• Hit architectural decisions that would be expensive to re-derive (e.g., "this repo uses the mock/real adapter pattern")
• Reach meaningful checkpoints on multi-session tasks (e.g., "we refactored 12 of 30 components; remaining 18 are...")
• Record explicit user corrections or preferences that should apply going forward

Without memory: Every session re-analyzes from scratch. Lessons learned are forgotten. The user repeats corrections. Multi-session work loses context and restarts.

With memory: A lightweight index points to relevant context. New sessions scan it before digging into code. User feedback is captured once and applied automatically. Multi-session tasks resume with a clear checkpoint.

Where Memory Lives