007-memory-learning

A project rule for managing durable agent memory and a learning counter through approved files and command-line tools.

In plain words
What is it for?
Use it when reading or updating project memory, syncing its generated cache, checking learning-module counts, and following module dependencies before execution.
Why use it?
It keeps long-term memory separate from generated caches and prevents direct edits to protected workflow state from corrupting project tracking.

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/phuoctrung-ppt/ai-sdlc-workflow/007-memory-learning
Clone the repo
git clone --depth 1 https://github.com/phuoctrung-ppt/ai-sdlc-workflow

Made for: Cursor.

Per session 346 This file is loaded in full into every session.
When invoked 346 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 $0.00346 $0.00346
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 7fb0875ab331, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

007-memory-learning 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.

.cursor/rules/007-memory-learning.mdc · 38 lines

What it actually says

Memory & Learning Rules

Memory

Location Role
docs/memory/* Primary durable SoT
.memory/* Generated AGENTS cache only (memory-loader.py --sync)

Read docs/memory/* first. Never hand-edit .memory/* as durable memory.

workflow-state.json — split concerns

Keys Owner Agent access
modulesSinceLastProposal, lastModuleCompleted, lastSkillProposal* Learning Only via python3 .cursor/scripts/learning-counter.py get|inc|reset
editedFiles, stopBlockCount, overrides, events workflow-guard.py Never in agent context

Forbidden: cat / open / hand-edit .cursor/state/workflow-state.json.

record-edit ignores .cursor/state/** and .aisdlc/** so counter updates do not pollute editedFiles.

Learning

  • Phase 6 /dev-module: distill memory + learning-counter.py inc --module <name> + lightweight @learning-agent
  • Full pass: /skill-update or when get returns fullPassRecommended: true
  • Proposal written → learning-counter.py reset --proposal <path>

Module deps

Before execute: read docs/module-deps.md; stop if upstream not done.

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 · 38 lines · 346 tokens per session scan A 7fb0875ab331

Subscribe to this mod's changes

007-memory-learning is a cursor rule published in the GitHub repository phuoctrung-ppt/ai-sdlc-workflow (2 stars, last pushed 16d ago), licensed MIT. It adds 346 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.