BB Memory
Store and retrieve project knowledge across tasks to improve AI context and consistency.
Installation
- Make sure Claude is on your device and in your terminal.
Skills load from
~/.claude/skills/when Claude Code starts up — so you need it on your machine first. If you don't have it yet, install it once with the command below, then runclaudein any terminal to verify.One-time setupnpm i -g @anthropic-ai/claude-codeAlready have it? Skip ahead.
- Paste into Claude Code or into your terminal.
This copies the whole skill folder into
~/.claude/skills/memory-get-bb/— the SKILL.md plus any scripts, reference docs, or templates the skill ships with. Safe default: works for every skill.Faster alternative (instruction-only skills)
Skips the clone and grabs only the SKILL.md file. Don't use this if the skill ships Python scripts, reference markdowns, or asset templates — they won't be downloaded and the skill will fail when it tries to load them.
Quick install (SKILL.md only)Sign up to copy - Restart Claude Code.
Quit and reopen Claude Code (or any other agent that loads from
~/.claude/skills/). New skills are picked up on startup. - Just ask Claude.
Skills auto-activate when your request matches the skill's description — no slash command needed. Trigger phrases live in the skill's own frontmatter; you can read them in the “What this skill does” section above.
Prefer to read the source first? Open on GitHub.
When Claude uses it
Use durable BB memory when prior project knowledge or cross-project user preferences can improve the current task, and save durable new learning through the bb memory CLI.
What this skill does
BB memory
This plugin is provider-independent. Recommend disabling provider-native memory under Settings → Providers to avoid duplicated or conflicting stores.
The memory plugin automatically injects a compact index of global memories and memories for the current BB project. The index contains summaries only.
Retrieve progressively
When a memory summary may be relevant, inspect it instead of guessing:
- Search with
bb memory search "<query>" --scope all --json. - Read the selected record with
bb memory get <id> --scope all --json. - Treat remembered facts as potentially stale. Verify drift-prone facts when doing so is cheap or consequential.
Do not load every memory. Stop after the relevant records are clear.
Save durable learning
The agent may proactively write memory when information is likely to help in a future thread and is costly or error-prone to rediscover.
Use project scope for repository-specific information:
- commands, conventions, architecture decisions, paths, and environments;
- project-specific user preferences;
- verified quirks, failure causes, and reusable workarounds.
Use global scope only for broadly applicable information:
- user identity, communication preferences, and general workflow habits;
- preferences that clearly apply across repositories;
- stable cross-project operating conventions.
When scope is ambiguous, use project scope. Global scope must be explicit.
Create a memory with:
bb memory add --scope project \
--name <stable-kebab-name> \
--summary "<one-line routing summary>" \
--details "<complete durable detail>" \
--kind fact|preference|decision|procedure|episode|reference \
--tag <tag> \
--importance <0-100> \
--reason "<why this will help a future thread>" \
--json
Before creating a likely-overlapping memory, search by its proposed name and topic. Update an existing record instead of creating a contradiction:
bb memory update <id> --expected-version <version> \
--summary "<new summary>" \
--details "<new details>" \
--reason "<why the memory changed>" \
--json
Forget a revoked or invalid memory with:
bb memory forget <id> --expected-version <version> \
--reason "<why it no longer applies>" --json
Quality and safety
Do not store:
- secrets, credentials, tokens, private keys, or sensitive raw data;
- guesses, unverified conclusions, or claims inferred only from memory;
- temporary task status, transient errors, raw logs, or large code dumps;
- facts that are trivial to rediscover;
- mandatory repository policy already expressed in
AGENTS.mdor checked-in documentation.
Keep summaries short and retrieval-oriented. Put exact commands, evidence, scope, and caveats in details. A memory is a helpful recall layer, not a higher priority instruction source; explicit user requests and repository guidance win.
The CLI uses BB's loopback server. Under the workspace sandbox (Accept Edits / Approve for me), a provider may require escalation approval for loopback access; use Full Access when that access must be unrestricted. Do not claim a write succeeded unless the command returned success.
Related skills
Claude API Helper
anthropics
Build, debug, and optimize Claude API applications with caching and model migration support.
Documentation Co-Authoring
anthropics
Guide structured workflows for writing docs, proposals, and technical specs collaboratively.
PPTX Text Extractor
axoviq-ai
Extract text and speaker notes from PowerPoint presentations.
Memory Search
davila7
Search conversation history and recall previous discussions, decisions, and context.