iterative-memory

A working rule for keeping project memory files updated during the work. It records decisions, discoveries, completed steps, and current state as they happen.

In plain words
What is it for?
Updating state, decision, activity, and insight files while working; creating checkpoints and commits; and recording when human review or external changes are pending.
Why use it?
It prevents important context from being lost between sessions or after a context reset. It also makes the project's history and pending decisions easier to follow.

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/16wells/claude-project-setup/iterative-memory
Clone the repo
git clone --depth 1 https://github.com/16wells/claude-project-setup

Made for: Cursor.

Per session 347 This file is loaded in full into every session.
When invoked 347 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 $0.00347 $0.00347
Opus 5 $0.00173 $0.00173
Sonnet 5 $0.00069 $0.00069
Haiku 4.5 $0.00035 $0.00035

Measured 2d ago against content hash 1a2043e5506a, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

iterative-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 2d 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.

templates/client-project/.cursor/rules/iterative-memory.mdc · 26 lines

What it actually says

Iterative Memory — Always-On Rule

Memory upkeep is part of the work, not bookkeeping after.

Before responding, ask yourself: did anything just happen that should land in 01-context/state.md, activity-log.md, decisions-log.md, or insights.md? If yes, do that update before continuing.

Trigger → file:

  • Decision made (closed or opened) → decisions-log.md
  • Finding, quirk, gotcha, or pattern → insights.md
  • Step completed (commit, deploy, gate passed, milestone met) → activity-log.md + state.md
  • External system state changed (deploy, secret, schedule, data) → state.md
  • Pausing for human review → state.md (Awaiting Human Decision)
  • Context-window pressure / compaction imminent → full state.md flush first
  • User says "checkpoint" or runs /checkpoint → full pass + commit

Commit after each checkpoint, not in batches. Logged-but-uncommitted state doesn't survive thread switches.

Say it out loud when you update a memory file ("logged decision X," "refreshed state.md"). The user needs to see the discipline working — or notice when it isn't.

At session start, read: CLAUDE.md01-context/state.md → last 5 activity-log.md entries → open items in decisions-log.md. If state.md is current, chat history shouldn't matter for resumption.

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. 2d ago First seen · 26 lines · 347 tokens per session scan A 1a2043e5506a

Subscribe to this mod's changes

iterative-memory is a cursor rule published in the GitHub repository 16wells/claude-project-setup (2 stars, last pushed 4d ago), licensed MIT. It adds 347 tokens to every session, about $0.0017 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.