What problem does it solve? AI assistants do not remember previous conversations between sessions, forcing users to re-explain preferences, past decisions, and recurring conventions every time. This Skill solves that with a single portable Markdown file — the project memory — that the user keeps alongside the project and attaches at the start of each new session, so durable context accumulates session after session under full user control. ## Core Features & Use Cases - Three-phase workflow: Load the memory file at session start, capture new preferences, lessons, and conventions during the conversation, then consolidate and version the updated file at the end or on demand. - Strict content boundaries: A 7-day test separates durable facts (preferences, decisions, lessons) from ephemeral task progress, which is routed to a project-state file instead of polluting the memory. - Security-aware consolidation: Facts from untrusted content are recorded with their origin attached, imperative instructions are rejected in favor of declarative facts, and no secrets or credentials are ever stored. - Use Case: At the end of a working session where the user corrected a recurring mistake and set a new reporting convention, ask for a memory consolidation and receive an updated, versioned MEMORIA-PROJETO.md to reattach next session. ## Quick Start Use the memoria-de-projeto skill to consolidate what we learned in this session into the project memory file and prepare the handoff for next time.
