All slash commands included in the memory engine. Each command has both English and Chinese versions — they do the same thing.
Type
/overviewor/全覽inside Claude Code to see this list.
Save something to memory so the next conversation knows about it. Checks for duplicates first — if the topic already exists in a memory file, it updates that section instead of creating a new one. Memory is stored in topic-specific files (memory/*.md), not dumped into one big file.
Load memory into the current conversation. SessionStart hooks inject a brief summary automatically, but /reload does a full read — loading actual file contents so Claude can reference specific details. You can also target a specific topic: /reload blog only loads blog-related memory.
Search across all memory files by keyword. Scans memory files, session diaries, learned skills, and session summaries. Case-insensitive, shows up to 20 results sorted by most recently modified.
Push memory files to your GitHub backup repo. The SessionEnd hook auto-commits locally after every conversation; this command handles the push to remote. Only pushes files that have actually changed.
Two-way sync: push local changes to GitHub, then pull any remote changes back. Use this when working across multiple devices. If there are conflicts, it shows them and asks how to resolve.
Disaster recovery and cross-device sync. When local memory is lost, corrupted, or you're setting up a new computer — this pulls everything back from your GitHub backup. Always pushes first (to protect current state), then pulls. With a GitHub backup repo, this works across devices: memories from your laptop can be pulled to your desktop.
Quick daily health check. Scans memory file structure, checks MEMORY.md capacity (200-line system limit), finds orphan files and broken links, and reports environment status. Lists suggestions but doesn't auto-modify anything.
Comprehensive weekly audit. Everything in /check plus: command layer audit (overlapping commands, broken paths), cross-file duplicate detection, cross-reference integrity (MEMORY.md -> memory files -> skills -> references chain), environment config check (settings.json, hook scripts), and git repo status across all projects.
Focused health check on memory files and hooks. Shows line counts, last-updated dates, and status for every memory file. Flags stale files (30+ days without update) and oversized files (200+ lines). Also checks hook script status via debug logs.
Auto-learn from mistakes. Claude triggers this on its own when it hits a pitfall — wrong approach that eventually got fixed, non-obvious workaround, or something that took 3+ attempts. Saves the problem, solution, and "next time" trigger to ~/.claude/skills/learned/. Stays quiet about it — just says "Learned: {title}" and keeps working.
The big review. Runs the Learning Loop's review steps: reads recent sessions, scans pitfall records, audits all memory files, marks each section as valid/outdated/duplicate, then suggests merges, updates, and cleanup. Uses a decision tree (serve the main goal? -> condense? -> already covered? -> only then delete). Generates a report with findings, a cleanup checklist for confirmation, and conclusions. Recommended weekly.
Write a reflection diary entry for the current conversation. Records what was done, what was learned, patterns noticed, and notes for next time. Saved to ~/.claude/sessions/diary/. Kept concise — under 30 lines per entry.
Cross-project task tracker. Shows all incomplete tasks grouped by project, suggests what to work on next based on urgency. Can filter by project name.
Helps decide when to use /compact. Shows a decision table: compact after research/debug/deploy/task-switch, don't compact mid-coding or mid-debug. Auto-suggests when context usage exceeds 60%.
This page. Lists all available commands with descriptions.