Borrowing it
Nothing to install: this file belongs to shahzebqazi/mhn-ai-agent-memory. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/shahzebqazi/mhn-ai-agent-memory/main/.cursor/skills/mhn-project-working-memory/SKILL.mdgit clone --depth 1 https://github.com/shahzebqazi/mhn-ai-agent-memoryWrote 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/shahzebqazi/mhn-ai-agent-memory/mhn-project-working-memory)<a href="https://agentmods.dev/skills/shahzebqazi/mhn-ai-agent-memory/mhn-project-working-memory"><img src="https://agentmods.dev/badge/skills/shahzebqazi/mhn-ai-agent-memory/mhn-project-working-memory/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/shahzebqazi/mhn-ai-agent-memory/mhn-project-working-memory"><img src="https://agentmods.dev/badge/skills/shahzebqazi/mhn-ai-agent-memory/mhn-project-working-memory.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00064 | $0.00682 |
| Opus 5 | $0.00032 | $0.00341 |
| Sonnet 5 | $0.00013 | $0.00136 |
| Haiku 4.5 | $0.00006 | $0.00068 |
Grade A, and why
mhn-project-working-memory scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 38 lines — stays where its author put it; the contents beside it link to each section on GitHub.
MHN project working memory (Cursor)
Idea
Associative memory lives in a JSON file on disk (not in chat). Any agent with the same MCP config reads and updates the same store, so memory is swappable between agents and survives new conversations.
Setup (once per machine)
- Create the MCP server venv: see
mcp-server/README.md(pip install -e .insidemcp-server/). - Copy
.cursor/mcp.json.exampleto.cursor/mcp.jsonin this repo (or merge thehopfield-memoryentry into user-level MCP config). - Replace placeholder paths with absolute paths to
mcp-server/.venv/bin/pythonandmcp-server/server.py, or use${workspaceFolder}if your Cursor build expands it. - Set
HOPFIELD_STATE_PATHto a path inside the project (default example:.mhn/working-memory.json). EnableHOPFIELD_AUTO_SAVE=truesostore/store_negativepersist immediately. - Restart MCP / Cursor so the server picks up env vars.
The bundled .cursor/mcp.json.example sets HOPFIELD_ENCODER to sentence_transformer; install the parent package with semantic extras into the same venv as the MCP server (./mcp-server/.venv/bin/pip install -e ".[semantic]" from repo root), or change that env var to random if you want no extra dependencies (weaker recall).
Use sentence_transformer (or better encoders) for semantic recall; random is fine only for exact-token overlap demos.
Agent workflow
working_memory_status— Confirmstate_path,auto_save, andnum_factsbefore relying on memory.list_facts— Treat as a cheap “table scan” of the local knowledge base (all stored strings).retrieve— Ranked facts with weights for a cue (like a fuzzy DB query).query_or_none— Prefer when the agent must distinguish “found in memory” vs “nothing relevant” (setmin_similarityif needed).store— Persist decisions, constraints, file paths, API shapes, user prefs, open questions. Keep facts short and atomic when possible.save/load— Optional explicit snapshots to other paths (e.g. backup or experiment branch).
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.
- 12d ago First seen · 38 lines · 64 tokens per session scan A 22bbc1ec7248
mhn-project-working-memory is a skill published in the GitHub repository shahzebqazi/mhn-ai-agent-memory (5 stars, last pushed 3mo ago), licensed MIT. It adds 64 tokens to every session and 682 once invoked, about $0.0003 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
pptx
A guide for turning a paper, outline, or structured text into a PowerPoint presentation. It covers an 8–12-slide structure, concise slide points, and generation of a .pptx file.
ai-style
A Chinese-language writing guide for producing or revising reader-facing text such as product announcements, public-account articles, emails, and README files. It focuses on reducing patterns that make writing sound machine-generated.
triage
A task-routing guide that identifies goals, dependencies, and acceptance conditions before work begins. It orders work as research, calculation or execution, and writing.
writing
A writing guide for turning verified facts and calculations into finished text for a specific audience. It follows the requested language, structure, and length.
data_analysis
A data-analysis guide for doing mathematical calculations and descriptive statistics with confirmed, sourced data. Descriptive statistics summarise data using measures such as totals, averages, or ranges.
coding
A coding guide for writing and running Python programs in a sandbox. It requires scripts to be small and reproducible, with their actual output or errors reported.