remember

To save coding conventions and best practices from a conversation, you can use a memory capture workflow that stores durable learnings into a persistent AGENTS.md file or generates a reusable skill structure for future reference.

Capturing architecture decisions is best handled by extracting them from discussions and storing the knowledge in project memory files like AGENTS.md, or by structuring them into a new reusable SKILL.md with proper frontmatter.

Storing conversation learnings as a reusable skill works by evaluating the identified workflows or guidelines, then automatically deciding whether to append the knowledge to global or project memory or generate a new structured SKILL.md.

Yes, this memory capture approach supports both global and project memory. It enforces clear placement of knowledge by allowing you to store durable learnings in either a broad global memory file or a localized project-specific AGENTS.md file.

Yes, you can automatically generate a new skill from a discussion about workflows. The process decides whether to save the learning to memory or create a new skill, structuring a SKILL.md with proper frontmatter and optional subdirectories.

You should save a learning to memory versus creating a new skill based on its scope: isolated guidelines or best practices go to AGENTS.md memory, while systematic, reusable workflows are structured into a new SKILL.md for repeated application.