Memory search

Skill S3YED/appie-kit/skills/misc/memory-search

Build Your Own AI Employee. The complete starter kit for OpenClaw + Hermes Agent. 155 deduplicated skills, drag-and-drop workspace, case studies, install scripts.

Install

npx -y skills add S3YED/appie-kit --skill memory-search

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

6 stars6 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Search and retrieve information from the Appie brain — daily logs, project notes, decisions, topics, and long-term memory files stored in ~/.hermes/memory/. Use this skill whenever you need to recall past conversations, decisions, project context, personal preferences, or anything discussed previously.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

3.8 KB, as published. Nobody here has run it

Memory Search

Search the Appie brain — daily logs, project notes, decisions, topics, and long-term memory.

When to Use

User asks about past decisions, conversations, or context
User says "remember when...", "what did we decide about...", "check my notes on..."
You need project context, client info, or infrastructure details
Looking up dates, events, or historical information

Memory Structure

All files are in ~/.hermes/memory/:

Daily logs: YYYY-MM-DD.md — raw daily notes
Topics: topics/*.md — subject-specific knowledge
Projects: projects/*.md — project documentation
Decisions: decisions/*.md — recorded decisions
Infrastructure: infrastructure/*.md — server/fleet notes
Plans: plans/*.md — future plans
Research: research/*.md — research notes

Top-level brain files in ~/.hermes/:

MEMORY.md — curated long-term memory (read this first)
USER.md — info about the human (Seyed)
TOOLS.md — tool configuration and fleet SSH details
IDENTITY.md — agent identity

Search Commands

Full-text search across all memory files

grep -ril "SEARCH_TERM" ~/.hermes/memory/ ~/.hermes/MEMORY.md ~/.hermes/USER.md ~/.hermes/TOOLS.md 2>/dev/null

Search with context (shows surrounding lines)

grep -rin -C 3 "SEARCH_TERM" ~/.hermes/memory/ 2>/dev/null | head -60

List recent daily logs

ls -t ~/.hermes/memory/????-??-??.md 2>/dev/null | head -10

Read a specific daily log

cat ~/.hermes/memory/YYYY-MM-DD.md

Search by date range (e.g., March 2026)

ls ~/.hermes/memory/2026-03-*.md 2>/dev/null

Search topics

ls ~/.hermes/memory/topics/ 2>/dev/null
grep -ril "SEARCH_TERM" ~/.hermes/memory/topics/ 2>/dev/null

Search projects

ls ~/.hermes/memory/projects/ 2>/dev/null
grep -ril "SEARCH_TERM" ~/.hermes/memory/projects/ 2>/dev/null

Advanced: fuzzy search with script

python3 ~/.hermes/skills/productivity/memory-search/scripts/search.py "QUERY" [--limit 10] [--recent 7]

Quick Reference

Always start with MEMORY.md for curated long-term context, then search daily logs for specifics.

Workflow

Read ~/.hermes/MEMORY.md for high-level context
If more detail needed, search with grep -ril across memory/
Read the specific files that match
Synthesize and answer

Project PRD / "last wishes" recall pattern

When the user asks whether you remember a PRD, brief, handoff, or their "last wishes":

Ground the answer before claiming memory. Search both long-term memory and the likely project directory for PRD, Handoff, README, Open decisions, TODO, wishes, and the project/client name.
Prefer canonical project artifacts over chat fragments: docs/PRD*.md, HANDOFF.md, README.md, AGENTS.md, PROJECT_STATE.md, and deployment manifests.
If multiple projects match, use the strongest local context and say which project you verified. Ask only if the ambiguity changes the answer.
Separate confirmed facts from inferred "last wishes". Example shape: "Verified PRD says X. Handoff/open decisions say Y. I do not see a separate file literally named last wishes."
Keep the reply short unless the user asks for the full doc. For Seyed, answer with the remembered decisions and blockers, not a tool-by-tool narrative.