Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/xxyoudeadpunkxx/decision-rain-library-project/llms.txt

Use this file to discover all available pages before exploring further.

This guide walks through everything you need to get Decision Rain Library Project running from scratch. By the end, you will have the five-collection Raindrop structure in place, the SYSTEM governance documents loaded as bookmarks, an AI assistant connected and oriented to the library, and at least one link processed through the full ingestion flow.

​
Prerequisites

Before starting, make sure you have the following:
  • A Raindrop.io account — the free tier is sufficient for this setup
  • Access to an AI assistant that can connect to Raindrop, such as:
    • ChatGPT with the Raindrop connector enabled (requires a ChatGPT account with connector access)
    • Another MCP-capable assistant configured with Raindrop access
No coding, automation, or self-hosted infrastructure is required. The entire system runs inside Raindrop and your AI assistant.

​
Step 1: Create the Collection Structure

1

Create the parent collection

In Raindrop, create a new collection at the top level and name it exactly:
Decision Rain Library Project
2

Create the five sub-collections

Inside the parent collection, create five nested sub-collections with these exact names:
00_SYSTEM
00_INBOX
10_REVIEW
20_LIBRARY
90_ARCHIVE
3

Verify the naming

Keep the names exactly as shown — including the numeric prefixes and underscores. The AI assistant will reference these collection names when saving and moving entries. Any variation will break the expected flow.

​
Step 2: Load SYSTEM Documents

SYSTEM documents are governance entries stored as Raindrop bookmarks inside the 00_SYSTEM collection. Because Raindrop requires every bookmark to have a URL, SYSTEM entries use stable placeholder URLs as addressable shells. The bookmark’s excerpt or note field — not the URL — is the source of truth for each document. Create one bookmark in 00_SYSTEM for each of the following placeholder URLs: 
https://example.com/rdl-agent-contract
https://example.com/rdl-system-spec
https://example.com/rdl-tag-registry
https://example.com/rdl-decision-rules
https://example.com/rdl-note-template
https://example.com/rdl-examples-golden
https://example.com/rdl-usage-flow
For each bookmark, paste the full content of the corresponding governance document from the docs/ folder of the Decision Rain Library Project repository into the bookmark’s note or excerpt field. The seven documents map to the seven placeholder URLs in the order shown above, from 00_AGENT_CONTRACT.md through 06_USAGE_FLOW.md.
If the AI assistant tries to open a placeholder URL and cannot read the Raindrop excerpt, it must stop and ask for access to the SYSTEM content. It should not reconstruct rules from memory, connector descriptions, or README fragments. If your assistant does this, check that it has read access to the 00_SYSTEM collection and that the note fields are populated.

​
Step 3: Connect Your AI Assistant

Connect the Raindrop MCP connector to your AI assistant following the connector’s current setup flow. The exact screens and steps change as platforms update their connector interfaces — treat the instructions in your assistant’s settings as the definitive UI guide, not this page. Once the connector is active, give it a short description that orients it to the Decision Rain Library Project operating model. Copy the following as your connector description: 
Use Raindrop as a Decision Library backend, not as generic bookmarks.
New links should first be saved to 00_INBOX.
Before reviewing or moving links, read 00_SYSTEM.
ChatGPT may inspect and propose, but the user decides final placement, tags, and next steps.
This connector description is a bootstrap instruction, not the full operating system. The canonical rules live in 00_SYSTEM so they can be inspected, versioned, and corrected. The connector description only needs to tell the assistant where to start and what authority it holds.

​
Step 4: Process Your First Link

1

Give the assistant a URL

Find a link you want to evaluate — a GitHub repository, a tool’s documentation page, an article, a product page, or any other external resource.
2

Ask the assistant to save and inspect it

Tell your assistant to save the link to 00_INBOX, inspect the page, and produce a review. A prompt like the following works well:
Save this link to 00_INBOX, read the page, and give me a review with a verdict, next step, claim, evidence, authority, truth assessment, and practical fit.
The assistant should first save the link to 00_INBOX to preserve the trace, then read 00_SYSTEM to orient itself, then inspect the source and produce the proposed note.
3

Review the proposal

Read the proposed verdict, tags, and next action. Consider whether the claim matches what the page actually describes, whether the evidence is credible, and whether the practical fit assessment matches your setup and constraints.
4

Approve, correct, or reject — then direct the move

Tell the assistant what to do with the entry. Approve the proposal and name the target collection, correct any tags or verdict before approving, or reject the entry and send it to 90_ARCHIVE. The assistant updates and moves the entry only after this explicit instruction.

​
What’s Next

Once your first link is processed, explore the rest of the documentation to get the most out of the system.

Daily Flow

The normal operating loop for processing new signals, reviewing pending entries, and maintaining the library over time.

Tag Registry

The controlled tag grammar — all approved families, allowed values, and rules for requesting additions.

Note Template

The canonical note format for evaluated entries, including the anti-truncation rule and the Claim/Evidence/Verdict invariant.

Agent Contract

The full authority boundary document: what the AI assistant may do, what it may not do, and what stop conditions apply.

Build docs developers (and LLMs) love

Get started for freeTalk to us