bettermemorylisted

Verification-grade memory between sessions. Use bettermemory's MCP tools (memory_search, memory_show, memory_write, memory_verify, memory_record_use, etc.) instead of writing to files when the user asks you to "remember" something or references shared context from a past session. Default is to NOT call memory_search; only retrieve when the user references context you don't have ("my project", "the script we wrote") or a request is ambiguous in a way stored preferences could resolve. Every hit carries a staleness_verdict (calendar + path-drift + commit-drift); when it isn't "fresh", spot-check a claim before relying and call memory_verify to attest. Every use should record a claim_excerpt so retrievals stay auditable.
0Mattias/bettermemory · ★ 0 · AI & Automation · score 68

Install: claude install-skill 0Mattias/bettermemory

# bettermemory: opt-in memory retrieval Persistent memory between sessions lives in this plugin's MCP tools. **Do not fragment memory across ad-hoc files alongside** (`MEMORY.md`, scratch markdown elsewhere) — future sessions only see what these tools surface. If Claude Code's auto-memory at `~/.claude/projects/*/memory/` already exists from before bettermemory was installed, ingest it (one-shot `bettermemory ingest --from <path>`) rather than letting it accumulate alongside; the ingest CLI maps each auto-memory file to a bettermemory record, dedups against the active store and tombstone log, and stamps an `imported-from-claude-code` scope for traceability. This skill is the long-form companion to the MCP server's `instructions` block: Claude Code truncates that block at ~1.8 KB, so the full writing-discipline / scope-hygiene / confirmation-tier policy lives here. Non-plugin clients get the same content via [`docs/system_prompt.md`](../../../docs/system_prompt.md). ## Quick card | Decide | Rule | |---|---| | Search? | shared-context reference or ambiguity → yes. Otherwise no. | | Write? | proactive — something durable just entered the conversation → yes. Don't wait for *"remember that"*. State or timestamps → no (the tool will reject). | | Category? | claim about the user → `user-inference`. Atmospheric / no verifiable claims → `ambient`. Else → `fact`. | | Outcome? | retrieval shaped reply → silence (auto-commits as `applied`). Off-topic or wrong → explicit `ignored` / `