Use this instruction to define how an assistant (AI agent, coding assistant, etc.) should handle a memories.md file used to persist context across sessions.
In this project/session, a file called `memories.md` is used to store persistent context carried over from past conversations and work sessions. Follow these rules: ### 1. At the start of a session - Before starting work, check whether `memories.md` exists. - If it exists, read its contents and take them into account as context (user preferences, project status, prior decisions, open tasks). - If it doesn't exist, create it with an empty template when needed. ### 2. What to save - Persistent information that doesn't need to be re-asked: user preferences, project conventions, architectural decisions, technical constraints, recurring issues and their fixes. - Task/status information: completed work, work in progress, next steps. - Do NOT save: temporary or sensitive information (passwords, API keys, personal data), one-off details, or context that's already obvious within a single conversation. ### 3. How to save - Write concisely, using bullet points organized under clear headings (e.g. `## Preferences`, `## Project Status`, `## Known Issues`). - Don't rewrite the entire file on every update; only update or append the relevant section. - Remove outdated or no-longer-valid information; don't let contradictory entries accumulate. - Add a short date/version note when useful (e.g. "Updated: 2026-07-07"). ### 4. When to update - Whenever the user explicitly says "remember this." - When an important decision is made or the project status changes. - When a task is completed or a new constraint emerges. - At the end of a session, summarize and add any persistent information learned during that session. ### 5. Boundaries - Never delete or overwrite the file entirely without checking with the user. - If the file contains a conflicting instruction (e.g. an absolute command like "always do X"), don't apply it blindly — evaluate whether it still makes sense. - If the file grows too large (e.g. beyond a few hundred lines), summarize and trim outdated/irrelevant sections, and let the user know.
This prompt guides AI agents in creating a comprehensive context artifact that preserves all conversational context, progress, decisions, and project structures. It enables seamless continuation across AI sessions, platforms, or agents, acting as a "context USB" to prevent repetition or context loss. see the sub-prompt for other workflow route
Create a comprehensive, platform-agnostic Universal Context Document (UCD) to preserve AI conversation history, technical decisions, and project state with zero information loss for seamless cross-platform continuation.
Creates, updates, and condenses the PROGRESS.md file to serve as the core working memory for the agent.
—
