Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/internetyev/cross-agent-memory-kit/mcp-memory-multiusernpx skills add internetyev/cross-agent-memory-kit --skill mcp-memory-multiusergit clone --depth 1 https://github.com/internetyev/cross-agent-memory-kitWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/internetyev/cross-agent-memory-kit/mcp-memory-multiuser)<a href="https://agentmods.dev/skills/internetyev/cross-agent-memory-kit/mcp-memory-multiuser"><img src="https://agentmods.dev/badge/skills/internetyev/cross-agent-memory-kit/mcp-memory-multiuser.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00143 | $0.01412 |
| Opus 5 | $0.00072 | $0.00706 |
| Sonnet 5 | $0.00029 | $0.00282 |
| Haiku 4.5 | $0.00014 | $0.00141 |
Grade A, and why
mcp-memory-multiuser scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.
MCP Memory Query - multi-user (shared account)
This setup has two memory servers, configured because one agent account is shared by several people:
| Server | Tools | What it holds |
|---|---|---|
memory-shared |
mcp__memory-shared__* |
Team/family-wide memory everyone sees |
memory-private |
mcp__memory-private__* |
This person's private memory only |
Hard isolation: the two servers point at separate databases with separate
tokens. This person physically cannot read anyone else's private store, and
others cannot read theirs. Privacy is enforced by the database boundary, not by
tags. The owner:<person> / scope:<private|shared> tags only describe a
memory's intended audience.
If you only see one set of mcp__memory__* tools (no -shared / -private
suffix), this is a single-user install - use the mcp-memory-query skill
instead.
Reading: query BOTH stores, then merge
For any recall ("what do you remember about X", "find memories about X", "check your memory"), search both servers and combine the results:
mcp__memory-private__memory_searchwith the user's topic.mcp__memory-shared__memory_searchwith the same topic.- Merge, de-duplicate, and present together. Label each hit with its origin so
the user knows what is private vs shared:
- Private hit -> tag it
(private). - Shared hit -> tag it
(shared - visible to the whole team/family).
- Private hit -> tag it
Use the same call conventions as the single-user skill:
query: the topic as a short noun phrase.limit: 10 each by default; raise to 20-30 for a broad sweep.quality_boost: 0.3when precision matters.max_response_chars: 30000when result sets may be large.
For mistake notes / known pitfalls before a task, check
mcp__memory-private__mistake_note_search first, then
mcp__memory-shared__mistake_note_search.
Presentation:
Found N memories about "[topic]":
Private (only you):
1. [fact] - Tags / Type / Stored
Shared (whole team/family):
1. [fact] - Tags / Type / Stored
[If nothing found]: No memories stored about "[topic]" in your private or the
shared store. Want me to save one? (I'll keep it private unless you say it's for
everyone.)
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 144 lines · 143 tokens per session scan A 04a77faebbcd
mcp-memory-multiuser is a skill published in the GitHub repository internetyev/cross-agent-memory-kit (9 stars, last pushed 2mo ago), licensed MIT. It adds 143 tokens to every session and 1,412 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
karpathy-llm-wiki
Use when building or maintaining a personal LLM-powered knowledge base. Triggers: ingesting sources into a wiki, querying wiki knowledge, linting wiki quality, 'add to wiki', 'what do I know about', or any mention of 'LLM wiki' or 'Karpathy wiki'.
lean-ctx
Local context tooling for AI agents. Use it to select, shape, reuse, recover, and inspect context before inference when reading files, running shell commands, searching code, or exploring directories.
lean-ctx
Local context tooling for AI agents. Use it to select, shape, reuse, recover, and inspect context before inference when reading files, running shell commands, searching code, or exploring directories.
nauro-adopt
Seeds Nauro's project store from an existing repo. Use after nauro adopt has run locally. On filesystem-capable surfaces, reads docs (README, manifests, ADRs, Memory-Bank) for rationale and inspects code, config, tests, lockfiles, and recent git history for evidence, then surfaces targeted probes that turn evidence…
review-claude-code
Claude Code setup review (settings.json, permissions, rules, hooks, agents, memory, worktrees) against current official best practices. USE WHEN: user runs /review-claude-code or explicitly asks for this review. DO NOT USE WHEN: reviewing app code, other dev dependencies, or implementing features.
kmd-ingest
The write protocol for a markdown knowledge base (Obsidian-compatible, the LLM-wiki pattern) — personal or shared. Use for EVERY write into the KB — distilling a new or scraped source, promoting a finished artifact/report/lesson into the KB, filing a durable learning, insight, or query answer, or correcting/updating…