source-integrity

A set of writing rules for checking claims against downloaded source files before using them in textbook chapters.

In plain words
What is it for?
It helps write, update, and edit textbook chapters with source verification, defined training-data limits, and rules for sub-agents.
Why use it?
It reduces confident errors such as wrong quotations, rounded statistics, and inaccurate descriptions copied from memory.

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/ai-learning-gems/ai-learning-gems.github.io/source-integrity
Clone the repo
git clone --depth 1 https://github.com/AI-Learning-Gems/AI-Learning-Gems.github.io

Made for: Cursor.

Per session 2,426 This file is loaded in full into every session.
When invoked 2,426 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 $0.02426 $0.02426
Opus 5 $0.01213 $0.01213
Sonnet 5 $0.00485 $0.00485
Haiku 4.5 $0.00243 $0.00243

Measured 2d ago against content hash 0464f8b442e1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

source-integrity 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/source-integrity.mdc · 155 lines

How it starts

The opening of the file, as written. The whole thing — 155 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Source Integrity Rules

Shared source verification and world-knowledge rules for all textbook chapter workflows (write, update, edit). These rules exist because LLMs confidently produce plausible-sounding but incorrect information about well-known topics. Every rule below was motivated by a real failure in a real chapter-writing session.


Why Source Integrity Matters

A textbook chapter that gets a quote slightly wrong, rounds a statistic, or describes a framework from memory instead of from its creator's actual words has silently corrupted the reader's knowledge. The reader trusts the textbook. They will cite the wrong number in their own paper. They will misattribute the quote. They will describe the framework incorrectly to their students. The corruption propagates.

The only way to prevent this is to treat every specific claim as unverified until it is confirmed against a downloaded, readable source file.


Zero World Knowledge Principle

You know nothing about the topic except what you read from the downloaded sources.

You are a skilled writer and organizer, but you have zero reliable knowledge about the chapter's topic. Your training data may contain information about the topic, but that information may be outdated, incomplete, or wrong. You MUST NOT:

  • Quote an author from memory (even a famous, widely-known quote)
  • Cite a statistic you "know" without reading the source
  • Describe a method, framework, or concept from training data instead of from a downloaded source
  • Fill in gaps when a source is unavailable by "remembering" the content
  • Assume a well-known fact is correct without verifying it in a source

If you cannot find a claim in a downloaded, readable source file, the claim does not exist for you. Drop it, or download a source that contains it.


What Training Data Can and Cannot Be Used For

The Zero World Knowledge Principle is strict, but not absolute. There is a precise boundary.

Hard Ban (Requires a Downloaded Source)

Read the full file on GitHub · 155 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. 2d ago First seen · 155 lines · 2,426 tokens per session scan A 0464f8b442e1

Subscribe to this mod's changes

source-integrity is a cursor rule published in the GitHub repository AI-Learning-Gems/AI-Learning-Gems.github.io (22 stars, last pushed 2mo ago), licensed MIT. It adds 2,426 tokens to every session, about $0.0121 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.