| name | evermemos |
| description | Search memories using EverMemOS. PROACTIVELY search before answering ANY project-related questions. Use when user asks about past conversations, previous decisions. ALWAYS check history before implementing features, debugging issues, or suggesting solutions. Maintain project continuity across sessions. Always put search results into your context window to improve your response. |
| argument-hint | search <query> [method] [top_k] |
| allowed-tools | Bash(python3 *) |
EverMemOS Memory Integration
Commands
The script to invoke is always:
~/.claude/skills/evermemos/scripts/evermemos_client.py
search
Search memories by query.
/evermemos search <query> [method] [top_k]
method: hybrid (default), agentictop_k: max results (default: 10)
When to use — ALWAYS trigger when:
- User asks "What did we discuss about X?" / "Did we fix that bug?" / "What approach did we decide on?"
- Any question containing: "last time", "before", "previously", "earlier", "remember"
- Before implementing features (check past approaches)
- Before debugging (check similar past issues and solutions)
- User mentions specific modules, files, or components → search that component first
- User asks how something works in THIS project
Execute:
For example, a default search:
python3 "$HOME/.claude/skills/evermemos/scripts/evermemos_client.py" search "<your search query>"
When more precise results or more results are needed, set method to agentic and top_k to a larger number (e.g. 15):
python3 "$HOME/.claude/skills/evermemos/scripts/evermemos_client.py" search "<your search query>" agentic 15
Proactive Usage Rules
Rule 1 — Search first, answer second (default behavior):
User asks a question
↓
Is it about THIS project or past conversations?
YES → SEARCH FIRST
NO → Answer directly
When in doubt, search. Missing context costs hours; an unnecessary search costs seconds.
Rule 2 — Multi-angle search for complex questions:
Don't search once and give up. Search multiple related angles. For example, if the user asks about authentication, then run the below:
python3 "$HOME/.claude/skills/evermemos/scripts/evermemos_client.py" search "authentication implementation"
python3 "$HOME/.claude/skills/evermemos/scripts/evermemos_client.py" search "auth bug fix"
python3 "$HOME/.claude/skills/evermemos/scripts/evermemos_client.py" search "auth security pattern"
Rule 3 — After getting search results:
Never search and ignore the results. Always use the search results as context to improve your response.
Retrieval Methods
Default: always use hybrid.
Only use agentic when:
- Query is vague or ambiguous (e.g. "that bug we discussed last month")
hybrid returned no useful results- User explicitly asks for deep or thorough search
Troubleshooting
Connection error: Check that EverMemOS is running (curl http://localhost:1995) and API_BASE_URL is correct.
No results: Try different keywords, switch to agentic method, or increase top_k.
