KNDL

Use this skill whenever you persist a fact to a memory filesystem, read facts back, or answer a question that requires recall from prior sessions. The skill stores each fact as a JSON-LD document with confidence, decay, bitemporal validity, provenance, and supersession, so an agent's memory can reason about what is fresh, what is stale, what is contradicted, and what came from where. Triggers: writing to /memory, reading from /memory, claims like "I learned that…", "I'll remember…", questions like "what did we decide…", "how confident are we in…", "who told us…", "is this still true". Do NOT trigger for ephemeral chat-only scratchpads or single-turn arithmetic. Pairs naturally with Anthropic's Memory on Managed Agents: KNDL is the file format inside the Memory store, this skill is the convention set Claude follows when reading/writing those files.

Use this skill whenever you persist a fact to a memory filesystem, read facts back, or answer a question that requires recall from prior sessions. The skill stores each fact as a JSON-LD document with confidence, decay, bitemporal validity, provenance, and supersession, so an agent's memory can reason about what is fresh, what is stale, what is contradicted, and what came from where. Triggers: writing to /memory, reading from /memory, claims like "I learned that…", "I'll remember…", questions like "what did we decide…", "how confident are we in…", "who told us…", "is this still true". Do NOT trigger for ephemeral chat-only scratchpads or single-turn arithmetic. Pairs naturally with Anthropic's Memory on Managed Agents: KNDL is the file format inside the Memory store, this skill is the convention set Claude follows when reading/writing those files.