What problem does it solve? In long-running AI-assisted development projects, the AI forgets prior requirements, decisions, and changes across sessions, leading to duplicated work, conflicting implementations, and repeated mistakes. This Skill builds a unified project memory so every downstream Skill starts with full historical context. ## Core Features & Use Cases - Project Memory Aggregation: Automatically reads PRD.md, DESIGN.md, PAGE_SPECS.md, DECISIONS.md, CHANGELOG.md, PROJECT_STATE.md, API_SPECS.md, and TASKS.md to build a task-relevant context summary. - Conflict & Risk Detection: Identifies duplicate development, decision conflicts (e.g., Redux vs Pinia), state conflicts, and API inconsistencies before code generation begins. - Context Compression: When project documentation exceeds 100 pages, generates a condensed Context Summary covering tech stack, page/component/API counts, and recent activity. - Use Case: Before asking an AI to add a batch-delete feature to a user management page, run this Skill to surface existing components (UserTable, SearchForm), existing APIs, the prior server-side pagination decision, and the risk that the batch-delete endpoint does not yet exist. ## Quick Start Ask the AI to run the context manager with your task description, such as "add batch delete to the user management page", before any code generation begins.
