Ben

Set up hmem-sync for cross-device memory synchronization. Use when the user wants to sync memories between devices, says "sync einrichten", "setup sync", "hmem-sync", or when /hmem-config detects hmem-sync is not installed. Covers first-device setup, additional-device restore, and MCP auto-sync configuration.

Curate hmem memory — your own or a foreign .hmem file. Systematically review entries, mark obsolete/irrelevant/favorite, fix titles, consolidate duplicates. Use when asked to "aufräumen", "clean up memory", "curate", "Speicher bereinigen", "tidy up", or when memory_health() shows issues.

Create a new P-entry. Uses the schema from hmem.config.json if configured, otherwise falls back to the standard R0009 schema. Use when the user says "neues Projekt", "new project", "Projekt anlegen", or "/hmem-new-project". Uses the create_project MCP tool which auto-creates P-entry (+ O-entry if createLinkedO is set).

View and change hmem memory settings, hooks, sync, and checkpoint configuration. Use this skill whenever the user types /hmem-config, asks to change memory settings, adjust parameters, tune bulk-read behavior, configure auto-checkpoints, manage hmem-sync, or troubleshoot memory-related issues. Also trigger when the user asks things like 'how often does auto-save fire', 'why is my context so large', 'change checkpoint to auto', 'how many tokens does startup cost', or 'set up sync'.

Load long-term memory from hmem. Use when: user types /hmem-read or says 'load memory', 'check your memory', 'what do you remember', 'show me the project', 'continue where we left off', 'was weißt du über...'; starting work with no L1 summaries in context; after /compact or context reset to reload knowledge; before significant work to anchor with prior context; user asks about a specific project, error, or topic visible in L1 summaries; user says 'pick up where we left off', 'resume', 'Woran haben wir gearbeitet', 'Was war der letzte Stand'. Calls read_memory() or load_project() immediately. Also covers search, prefix filter, context_for, stale detection, find_related, memory_stats, memory_health.

Store facts, preferences, decisions, and project context into hmem long-term memory using the write_memory MCP tool. Use when the user says 'remember this', 'save this', 'don't forget', 'store this for later', or invokes /hmem-write. Persists key lessons, error resolutions, architecture decisions, user preferences, and project state across sessions. Use when Claude should record conversation insights, save project context, persist important facts, or store user preferences for future reference.

Post-update checklist for hmem-mcp and hmem-sync. Run after npm update or when hmem detects a version change. Covers skill sync, entry migration, schema enforcement, O-entry curation, and smoke tests. Use when the user says 'update hmem', 'hmem updaten', or when the startup version-check detects a new version.

Set up and configure the hmem memory system. Run this skill to install hmem, initialize the MCP server, deploy skill files, and register auto-memory hooks for Claude Code, Gemini CLI, or OpenCode. Covers first-time setup ('set up memory', 'initialize hmem', 'first-time setup', 'install hmem'), manual installation, and post-setup verification.

Migrate O-entries to the new project-bound 5-level structure (v5.3+). Each O-entry gets permanently bound to its P-entry by matching sequence numbers (O0048 <-> P0048). Run this once after updating to v5.3. Use when the user says "migrate O-entries", "O-Eintraege migrieren", "hmem migrate", or when hmem-update detects that the O-entry structure needs migration. Also use proactively if you notice O-entries are still using the old flat format (L2->L4->L5 without L3 batches).

Prepare for /clear by optionally saving high-value knowledge. Use when: - User types /wipe - Context threshold warning appears (100k tokens) - User says "context aufräumen", "clear machen", "wipe" Handles pre-clear cleanup, then instructs user to /clear for automatic context restoration.

Search hmem long-term memory for something the user mentioned earlier but doesn't fully remember. Use whenever the user refers to past conversations without giving an ID — phrases like "wir haben letzte Woche/gestern/neulich/damals besprochen", "erinnerst du dich an", "was war nochmal mit", "wie hieß der Fehler von vorgestern", "vor ein paar Tagen hatten wir", or English equivalents ("remember when", "we talked about", "that thing from yesterday/last week"). ALSO trigger when the user references a topic, person, project detail, bug, or decision as if you should already know it — proper names you've never seen this session, definite articles ("der Bug mit X", "das Problem von neulich", "the issue we had"), or casual asides that presuppose shared context. If you catch yourself thinking "the user seems to assume I know this" — search first, ask second. Combines full-text search with the implied time window from the user's phrasing. Trigger even if the user sounds casual or vague — that's exactly when they need you to search rather than guess. Do NOT trigger for explicit ID lookups like "read P0048" or "show me L0042".