From mempalace
Initialize and set up mempalace — a local AI memory system — for the current project. Use this skill when the user mentions mempalace, wants persistent memory across Claude sessions, asks about remembering context between conversations, wants to set up project memory, or runs /mempalace. Also use when the user asks to mine conversations, search memories, or configure memory hooks. Even if they just say "set up memory" or "remember things between sessions", this skill applies.
npx claudepluginhub sadwx/mempalace-plugin --plugin mempalaceThis skill is limited to using the following tools:
Guides Payload CMS config (payload.config.ts), collections, fields, hooks, access control, APIs. Debugs validation errors, security, relationships, queries, transactions, hook behavior.
Designs, audits, and improves analytics tracking systems using Signal Quality Index for reliable, decision-ready data in marketing, product, and growth.
Enforces A/B test setup with gates for hypothesis locking, metrics definition, sample size calculation, assumptions checks, and execution readiness before implementation.
Fully automated setup of mempalace — a local AI memory system (ChromaDB + SQLite). This skill runs end-to-end with zero user interaction. Execute every step, report what you did at the end.
The plugin already handles MCP server registration and hook scripts automatically. This skill handles project-specific initialization: palace init, wing config, identity, and initial mining.
Before starting, ensure mempalace is installed:
bash ${CLAUDE_PLUGIN_ROOT}/scripts/ensure-installed.sh
If this fails, tell the user to install manually (pip install mempalace or uv pip install mempalace) and stop.
Verify:
mempalace --version
Run all steps sequentially. Skip any step whose check passes. Do NOT ask for confirmation — just do it.
Gather info automatically — no user input needed:
- and . with _, prefix with wing_ (e.g., psg-mcp becomes wing_psg_mcp)src/, tests/, docs/)mempalace init "$(pwd)"
If palace already exists, skip — verify with mempalace status.
Check where the wing config lives:
mempalace status
Write/merge wing config (typically ~/.mempalace/wings.json or inside the palace directory):
{
"default_wing": "wing_<project_name>",
"wings": {
"wing_<project_name>": {
"type": "project",
"keywords": ["<project_name>", "<derived_keywords>"]
}
}
}
If the config already exists with other wings, READ it first and merge — append the new wing, preserve existing wings. Only set default_wing if this is the first wing.
If ~/.mempalace/identity.txt does not exist, create it from git config and project context:
git config user.name
Write a ~50 token identity like:
<Name> — software engineer working on <project description>.
If identity.txt already exists, skip.
Seed the palace with current project context:
mempalace mine "$(pwd)"
Idempotent — safe to run even if already mined.
mempalace status
Read the plugin version from ${CLAUDE_PLUGIN_ROOT}/.claude-plugin/plugin.json (the version field).
Print a single summary:
Mempalace plugin v<version> — setup complete:
- Palace: initialized at <path>
- Wing: wing_<project_name> (keywords: ...)
- Identity: <first line>
- Mined: <project_path>
- MCP: auto-registered by plugin
- Hooks: Stop + PreCompact auto-configured by plugin
After setup, mempalace works automatically via MCP tools in Claude sessions. For manual use:
| Command | Purpose |
|---|---|
mempalace search "query" | Semantic search across memories |
mempalace search "query" --wing name | Search within a specific wing |
mempalace mine <path> | Ingest code/docs |
mempalace mine <path> --mode convos | Ingest conversation exports |
mempalace wake-up | ~170 tokens of critical context for session start |
mempalace status | Palace overview |
Run /mempalace in any project directory. The skill auto-detects the new project, creates a new wing, and wires everything up. Existing palace data and other wings are preserved.