memory-bank

memory-bank is a cursor rule for coding agents from pmikutel/directed-memory-bank. It costs 21 tokens per session (1,430 once invoked), scanned A, original, MIT.

A project knowledge-base workflow for memory-bank/, a folder that stores shared notes about ongoing and completed work. It guides an agent to check current tasks and load rules relevant to the area being changed.

In plain words
What is it for?
Use it when starting work, navigating an unfamiliar part of a project, or maintaining the memory-bank notes.
Why use it?
It helps an agent recover project context and follow the right local instructions when entering unfamiliar work.

Cursor rule

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/pmikutel/directed-memory-bank/memory-bank
Clone the repo
git clone --depth 1 https://github.com/pmikutel/directed-memory-bank

Wrote 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.

agentmods badge for memory-bank

README.md
[![agentmods](https://agentmods.dev/badge/rules/pmikutel/directed-memory-bank/memory-bank.svg)](https://agentmods.dev/rules/pmikutel/directed-memory-bank/memory-bank)
Your own site
<a href="https://agentmods.dev/rules/pmikutel/directed-memory-bank/memory-bank"><img src="https://agentmods.dev/badge/rules/pmikutel/directed-memory-bank/memory-bank.svg" alt="Measured on agentmods" height="20"></a>
Per session 21 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,430 The whole file, excluding the scripts and references it only reads on demand.
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 $0.00021 $0.01430
Opus 5 $0.00010 $0.00715
Sonnet 5 $0.00004 $0.00286
Haiku 4.5 $0.00002 $0.00143

Measured 5d ago against content hash 9006a5539293, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

memory-bank 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 5d 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.

template/integrations/cursor/memory-bank.mdc · 109 lines

How it starts

The opening of the file, as written. The whole thing — 109 lines — stays where its author put it; the contents beside it link to each section on GitHub.

DMB — Cursor entry

DMB is the project's shared knowledge base. This is an Agent-Requested rule — Cursor loads it on demand when the description matches the operation at hand. Description-only frontmatter (no alwaysApply, no globs:) keeps it cheap; the always-on essentials.mdc carries anything that must always be in scope.

Workflow — where to start

Step 1: glance at current work

The memory-bank/tasks/ layer is an optional scratchpad — it may or may not contain relevant context for what you're about to do.

  • memory-bank/tasks/work/ — one file per in-flight topic (idea, plan, active work). May be empty. If nothing relevant exists, that's fine.
  • memory-bank/tasks/log/ — one file per completed unit of work (typically one per merged PR). Useful for history / debugging context.
  • memory-bank/tasks/process.md — the workflow for maintaining work/ and log/.

Do not fabricate entries. If there's nothing in work/ that matches, move on.

Step 2: let path-scoped rules fire

Operation-specific routing lives in the per-area rules under .cursor/rules/ (backend, frontend, migrations, etc.). Each one names the memory-bank files relevant for that area and fires automatically when the editor is touching matching files. Trust them; do not duplicate their work here.

Step 3: navigate when you need to

If you need a file that no path-scoped rule covers — or you're not sure what knowledge the project has captured — read memory-bank/_index.md. It is the inventory of what exists and which files matter for which kind of work. Use it to discover what to load, not as a step taken every session.

The tasks/ layer

What it is

RAM memory for code work. It does not replace Notion, Jira, or any tool that defines what work exists.

work/ — optional scratchpad

One file per topic, human-readable kebab-case slug (e.g., choppy-audio-issue.md). Frontmatter carries state (status, branch, pr, started, updated). A topic's file grows from "idea" through "planned" / "active" and moves to log/ on PR merge.

Read the full file on GitHub · 109 lines

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. 5d ago First seen · 109 lines · 21 tokens per session scan A 9006a5539293

Subscribe to this mod's changes

memory-bank is a cursor rule published in the GitHub repository pmikutel/directed-memory-bank (5 stars, last pushed 1mo ago), licensed MIT. It adds 21 tokens to every session and 1,430 once invoked, about $0.0001 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.