agent-magnet

A set of conversation rules for connecting an agent to a remote memory service. It tells the agent to load stored context at the start, record user signals during the conversation, and save the session at the end.

In plain words
What is it for?
Use it to load and save memory, and to record likes, dislikes, corrections, rejected suggestions, and sensitive past experiences.
Why use it?
It helps preserve context and user preferences between conversations, so the agent does not need to start from scratch each time.

Cursor rule for Cursor

Install

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.

agentmods
npx agentmods add rules/helinakdogan/agentmagnet/agent-magnet
Clone the repo
git clone --depth 1 https://github.com/helinakdogan/agentmagnet

Made for: Cursor.

Per session 251 This file is loaded in full into every session.
When invoked 251 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00251 $0.00251
Opus 5 $0.00125 $0.00125
Sonnet 5 $0.00050 $0.00050
Haiku 4.5 $0.00025 $0.00025

Measured 3d ago against content hash ba89238406bd, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agent-magnet 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 3d 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.

.cursor/rules/agent-magnet.mdc · 19 lines

What it actually says

Memory Rules

At the start of every conversation, call inject_memory with user_id "helin" and project_id "default". Prepend the returned injection to your context silently — do not announce it.

While conversing, watch for behavioral signals:

  • User expresses a like or preference → call add_signal with user_id "helin", signal_type preference_like
  • User expresses dislike or says "don't", "never", "hate" → call add_signal with user_id "helin", signal_type preference_dislike
  • User corrects your output → call add_signal with user_id "helin", signal_type correction
  • User rejects a suggestion → call add_signal with user_id "helin", signal_type rejection
  • User mentions something sensitive or a past bad experience → call add_signal with user_id "helin", signal_type watch_out

Call these tools silently in the background. Never say "I'm recording this to memory" — just do it and behave accordingly.

At the end of the conversation, call save_session with user_id "helin" and the full message history.

Changes

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.

  1. 3d ago First seen · 19 lines · 251 tokens per session scan A ba89238406bd

Subscribe to this mod's changes

agent-magnet is a cursor rule published in the GitHub repository helinakdogan/agentmagnet (14 stars, last pushed 8d ago), licensed MIT. It adds 251 tokens to every session, about $0.0013 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 2026-08-30.