breadcrumb

To persist discovered tools and patterns for future Claude sessions, you register them as compact notes in the correct discovery layer, such as CLAUDE.md or MEMORY.md. This ensures future instances find the verified capabilities without rediscovering them.

The best way to save workflow memory and debugging gotchas is classifying the discovery and adding a short, reusable note to the appropriate persistent layer. This prevents duplicating full documentation while keeping the knowledge visible across sessions.

You record verified CLI flags by capturing them as a breadcrumb in the most appropriate discovery layer. This guarantees the next Claude instance will see the verified workaround and apply it consistently.

You should write workflow discoveries to CLAUDE.md versus MEMORY.md based on layer classification. The skill determines whether a breadcrumb belongs in CLAUDE.md, MEMORY.md, a specific Skill's SKILL.md, or a deeper reference document to ensure careful placement.

Yes, you can use this approach to avoid duplicating existing documentation entries. The skill requires compact note formatting and careful placement to ensure newly registered capabilities do not duplicate existing records in the knowledge base.