Skip to content
Agentmart
Submit a tool

memory

Use when starting, resuming, or ending non-trivial coding, debugging, RCA, incident analysis, or architecture/design work where prior failures, decisions…

Skill by Nomadcxx

IndexedBSD-3-Clause★ 1
Write a review

Chalk Labs

Need help deploying memory?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/nomadcxx-memory

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Use when starting, resuming, or ending non-trivial coding, debugging, RCA, incident analysis, or architecture/design work where prior failures, decisions, conventions, constraints, or durable user or team preferences may matter across sessions and should be recalled from or stored to the MCP-backed hybrid-memory system

Use cases: design, agent-building

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/memory

What's inside

Skill name
memory
Description
Use when starting, resuming, or ending non-trivial coding, debugging, RCA, incident analysis, or architecture/design work where prior failures, decisions, conventions, constraints, or durable user or team preferences may matter across sessions and should be recalled from or stored to the MCP-backed hybrid-memory system
License
BSD-3-Clause
Instructions
160 lines in SKILL.md · view on GitHub

Files (4)

  • SKILL.md
  • harvest.md · references/
  • recall.md · references/
  • triggers.md · references/

SKILL.md (first lines)

# Hybrid Memory Ops

## Overview

Use this skill to operate smolbot's MCP-backed `hybrid-memory` integration well.

The core rule is simple: recall small, retrieve in stages, and store only distilled durable leverage.

smolbot exposes the memory server through wrapped MCP tools:

- `mcp_hybrid-memory_memory_search`
- `mcp_hybrid-memory_memory_semantic`
- `mcp_hybrid-memory_memory_store`
- `mcp_hybrid-memory_memory_get`

Additional maintenance tools may also be available:

- `mcp_hybrid-memory_memory_delete`
- `mcp_hybrid-memory_memory_stats`
- `mcp_hybrid-memory_memory_cleanup`

## When To Use

Use this skill when:

- starting or resuming non-trivial coding, debugging, RCA, incident triage, or architecture/design work
- recurring bugs, repeated dead ends, or "we've seen this before" signals suggest prior debugging may matter
- a root cause, failed experiment, benchmark, mitigation, or rejected approach should be checked before repeating work
- making, revisiting, documenting, or explaining an architecture/design decision, tradeoff, or prior rationale
- stable user or team preferences, conventions, constraints, or operating habits are shaping implementation choices
- ending a session that produced durable causes, decisions, benchmarks, conventions, constraints, or preferences worth preserving

Do not use this skill to:

- treat small mechanical tasks as memory-worthy by default
- dump raw session notes into long-term memory
- treat `hybrid-memory` like a transcript archive
- run broad memory searches continuously

## Required Workflow

Always follow this 5-stage workflow:

1. decide whether startup recall is warranted
2. if warranted, do one small startup recall
3. do targeted mid-session lookup when recurring failures appear, an RCA question arises, an architecture/design choice is being revisited, or a durable preference/constraint may change the approach
4. run a brief harvest gate; a candidate passes only if all are true: it will likely help in a future session, it is specific enough to change behavior, it remains true beyond the current task or run, and it can be distilled into a short reusable statement
5. store only candidates that passed the gate and are new, corrective, or materially better than existing memory

Invoking this skill does not automatically require touching the memory backend.
If the startup gate is not met, skip recall.
If no durable outcome emerged, skip store.

## Startup Gate

At the start of a session, ask:

- is this a known repo or recurring system?
- is there likely prior debugging, architecture, or convention context that would materially change the next action?
- can I formulate a narrow project-scoped query instead of a broad search?

If the answer is no, skip recall and work normally.

If the answer is yes, prefer one small shortlist over a memory dump.

## Retrieval Order

Prefer this order:

1. `mcp_hybrid-memory_memory_search` for precise keyword lookups
2. `mcp_hybrid-memory_memory_semantic` only if keyword search is weak or ambiguous
3. `mcp_hybrid-memory_memory_get` only for shortlisted items that need detail

Good startup recall output:

- a compact shortlist
- 2-5 actionable bullets at most
- emphasis on decisions, errors, patterns, and conventions

Bad startup recall output:

- raw rows
- long dumps
- every matching result

## Mid-Session Triggers

Run targeted lookup when:

- a bug or failure pattern repeats
- you suspect "we solved this before"
- a project-specific convention likely exists
- a user preference materially affects execution
- an RCA or architecture decision needs prior context
- a benchmark or prior result may change next steps

Avoid lookup when:

- the task is routine and local
- there is no sign that prior context matters
- the query would be broad and noisy
- you are just filling silence with extra actions

## Harvest Gate

Before storing anything, ask whether the session produced one or more of:

- a validated root cause
- a durable decision with rationale
- a reusable pattern
- a durable project convention
- a user preference that changes future execution
- a benchmark or comparison result that should influence future work

If none of these happened, store nothing.

## Distill Before Store

Before calling `mcp_hybrid-memory_memory_store`:

From Nomadcxx/smolbot, BSD-3-Clause.

More from Nomadcxx/smolbot

Signals

Stars
1
Forks
0
Open issues
0
Last commit (this folder)
Apr 5, 2026 (6 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
BSD-3-Clause
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from Nomadcxx

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get memory running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying memory?

CallGet help
Cite this page

memory skill. Agent Mart. https://agentmart.dev/skills/nomadcxx-memory (last updated 2026-10-24)