• about photosynthesis and how plants communicate
• via colleague

• The list of MCP/GitHub/servers either in the src folder or in the root readme (which reads a bit like an AwesomeList) contain one.

• GitHub/Docs/Authentication/Account security/Manage personal access tokens
  • GitHub recommends that you use GitHub Fine-Grained PATs instead of GitHub Personal Access Token (Classic) whenever possible.
  • GitHub Fine-Grained PAT
  • GitHub PAT (Classic)

Security/Quest/Zero Secrets on Disk

• Resumed 1Password/Dev/How To/Replace .env secrets with op refs/Q/Can mise wrap op run
  • Reviewed mise/GitHub/Issue/24/02/Struggles with integrating 1Password op in greater detail while trying to use
    • #Discovered that the poster Person/Randy Syring uses or developed level12/GitHub/env-config which isn't what I want

      • env-config (EC) helps manage environment variables in the active shell as defined by an env-config.yaml configuration file. It has built-in support for reading secrets from 1Password.
        

  • #Filed
    • 1Password/Dev/Doc/Guide/Load Secrets/Into Scripts lists four options for loading secrets into scripts
      • My Note - #Important - while this page recommends using service accounts over using 1Password/Dev/op, only some people can use them
      • 1Password/Dev/CLI/Shell Plugin

        • • Summary

            • With 1Password Shell Plugins, you can configure 1Password to securely authenticate third-party CLIs with your fingerprint, Apple Watch, or system authentication. Your CLI credentials are stored in your 1Password account, so you never have to manually enter your credentials or store them in plaintext.
            • You can test shell plugins or choose a shell plugin from the list below to get started.
  • #Updated
    • 1Password Service Account

      • • Important limitations

          • While it sounds like a great idea, and 1P recommends Service Accounts over using 1Password/Dev/op, it's not available to all users or teams and has important use cases that aren't covered yet

          • Not a default feature of 1Password

            • According to 1Password/Dev/Doc/service-accounts/Manage/creating service accounts, only 1Password/Business admins can enable a group to create and manage service accounts. So you can use service accounts if you are a 1Password/Teams admin or if you in a 1Password Business setting in a Group that can create service accounts. 1Password/Family doesn't have any ability to use service accounts.
              • • 1Password/Dev/Doc/service-accounts/Manage/creating service accounts
                  • By default, only owners and administrators can create and manage service accounts in 1Password/Teams and 1Password/Business.
                  • To allow other groups to create service accounts, an owner or administrator can allow certain 1Password/Business/Groups to be able to manage service accounts.
              • to fully embrace Security/Service Accounts with fine-grained permissions, a team would need to create many service users, preferably with short-lived access credentials.

          • Can't programmatically delete short-lived service accounts

            • 1Password/Slack/cli message here from Vladyslav Ponoiko

              • My issue mainly appears because I need to use SA tokens, which I usually generate before run of a particular script and they're aimed onto liveness of max 1 day.
                Today I run into error when tried to create SA, so cleaned up couple of SA's and this worked, but I would love to get more consistent onto the 'cleaning up' part, as I have lots of SA's by now, and would really love to keep my way of doing things
                

              • response from 1Password jr software developer Person/Eddy Filip says that 1Password/Dev/CLI doesn't enable deleting service accounts for the short-lived case

                • currently the CLI doesn’t provide a programmatic way of deleting service accounts (either active or expired).
                  The only way you can delete service accounts is through the browser. The expired service accounts will be grayed out and will also show with red that they are expired.
                  In the meantime, I will file an issue internally with the request of adding support for deleting service accounts via the CLI.
                  

              • the conversation also references Service account management · Issue #115 · 1Password/onepassword-sdk-js

• watched Person/Brian Madison/YouTube/25/02/Unlock Cursor AI Automatic Rule Generation

• they have created a sample .mdc that documents mdc file structure and usage at https://github.com/justdoinc/justdo/blob/master/.cursor/rules/999-mdc-format.mdc

• • My Notes
    • I don't quite get it. Somehow they are claiming that this Project/Management tool is better for AI Coding because the source code is available, and therefore the AI can modify it to your needs.
    • They have cursor rules in JustDo/GitHub/justdo here - justdo/.cursor/rules at master · justdoinc/justdo - so clearly they are trying to use it this way ... but I don't quite understand what they are thinking about how a downstream repo is supposed to use cursor to customize this. Are they expecting that we'll fork it and use the built-in rules ... ? Also, what exactly is the integration ... is it able to use MCP to integrate with JustDo?
      • FWIW it's worth, their CursorAI/Project Rules are well organized

• git/reset
• git/How To/Discard Changes to Working Dir

• The github terraform provider is powerful, but not powerful enough to enable provisioning GitHub Personal Access Token

• • My Notes on trying this workflow - 2025-03-21 Fri

    • I prepared a Merge Request Fix spelling errors by codekiln · Pull Request #27 · bmadcode/cursor-auto-rules-agile-workflow into this repo to get to know the process. Here are my notes on the process. Similar to AI/Coding/Technique/Phased Planning/ai-coding dir, this repo by Person/Brian Madison presents a system of CursorAI/Project Rules that implement an Agile workflow.

      • Workflow Review

        • Onboarding

          • I was initially a bit confused by the introductory material in the root README.md, which contains important notices to new users before explaining what the repository does.
            • Note this does contain a great #Tip CursorAI/How To/update workbench.editorAssociations for mdc to prevent custom editor
          • Just skip down to the Overview for more information.

        • Installation

          • The root README.md presents two quick start options, **BOTH** of which rely on these installation instructions:
            • git clone https://github.com/bmadcode/cursor-auto-rules-agile-workflow.git
            • cd cursor-auto-rules-agile-workflow
            • ./apply-rules.sh /path/to/your/project
          • The result of running this against your project:
            • 1. Copy all template rules to your project's .cursor/rules/ directory

              1. Add documentation to docs/workflow-rules.md

              2. Update .gitignore to protect "private" rules (individual developer cursor rules that begin with _ by this project's conventions)

              3. Preserve any existing rules in your project

          • After running ./apply-rules.sh, the script will suggest next steps

            • Next steps:
              

            • 1. Review the documentation in docs/workflow-rules.md

              • note, in the original repository, in the docs/ dir there is only agile-readme.md, but after running ./apply-rules, workflow-rules.md does exist in the target repository. This file is actually hard-coded in apply-rules.sh!

            • 1. Choose your preferred workflow approach

              • for example, cursor rules (recommended) vs cursor notepads

            • 1. Enable Cursor Notepads if using the flexible workflow option

            • 1. To start a new project, use xnotes/project-idea-prompt.md as a template

              to craft your initial message to the AI agent
              

          • The generated docs/workflow-rules.md in the target repository recommends
            • using 801-workflow-agile.mdc, which is in the xnotes directory
              • Since this is not the official project rules directory, one needs either to reference it directly, or to copy it into the .cursor/rules directory
            • using 901-prd.mdc, 902-arch.mdc, 903-story.mdc, which are not in the repository anywhere AFAICT
              • It's possible these are subsumed into the three equivalent-ish templates in .cursor/templates, but that's not totally clear to me

        • Overview

          • This rule system has a few components. The two most highly visible in the core README.md are:

            • Using the agent to author and update custom CursorAI/Project Rules

              • similar to Person/Geoffrey Huntley/Blog/25/02/You are using Cursor AI incorrectly..., this repo presents a method of creating new cursor rules.
              • Each type of CursorAI/Project Rule has a template in .cursor/templatess/*.md
                • #Example #PRD template - cursor-auto-rules-agile-workflow/.cursor/templates/template-prd.md at main
                • I like this method of standardizing how cursor creates project rules by using configurable templates. This seems flexible and powerful.

            • Conventions for organizing cursor project rules using a numerical catalog and separating private and public rules

              • This reminds me of the Dewey Decimal System
                • 0XX: Core rules and standards
                • 1XX: Tool and MCP rules
                • 3XX: Testing standards
                • 8XX: Workflow rules
                • 9XX: Templates
                • 1XXX: Language-specific rules
                • 2XXX: Framework/library rules
              • I believe I remember reading something in the CursorAI/Forum about people who tested to see if the agent prioritizes items in numerical order, but I can't seem to find the link.
          • The project structure includes a docs dir with a single file - docs/agile-readme.md. This is the Agile part that's most like my AI/Coding/Technique/Phased Planning/ai-coding dir.
            • In my MR I used it or invoked it like this:

              • I've just forked this repository and checked out a branch. I've noticed there are a lot of spelling errors. I'd like to use this as an opportunity to try out the agile workflow mentioned in @agile-readme.md. Can you suggest what my first step should be to start an Epic to fix spelling and grammar and Stories for each file?
                

              • Note: in retrospect, I probably should have copied 801-workflow-agile.mdc to my cursor project rules and used that ?
            • The project rules guided the agent to create a directory structure like this:
              • .ai
                • a #PRD document prd.md, possible equivalent of AI/Coding/v0/File/1-TICKET-original-issue.md in my workflow
                • an arch.md, possible equivalent of AI/Coding/v0/File/2-TICKET-technical-specification.md which, for my pull request, contained an LLM-generated Mermaid Diagram that looked like this:
                  • 
                  • Note: I needed to install VSCode/Extension/Markdown Preview Mermaid Support to see it in a preview locally
                  • I #❤️ the idea of using mermaid in an Arch/itecture document
              • epic-1/
                • story-1.story.md
                • story-2.story.md
                • story-3.story.md
                • story-4.story.md

            • #Thoughts about the .ai/epic-<N>/story-<M>.story.md structure vs what I've been doing in AI/Coding/Technique/Phased Planning/ai-coding dir

              • I like the idea of keeping the stories as separate files. This part is different than the current version of my workflow; the prompts I've been using from Person/Harper Reed/Blog/25/02/My LLM codegen workflow atm end up producing two documents, each of which are broken up into 3-10 equivalent stages, each of which will end with a commit:
                • AI/Coding/v0/File/3-TICKET-staged-implementation-plan.md
                • AI/Coding/v0/File/4-TICKET-staged-implementation-checklist.md
              • Comparison of putting stages into separate files vs into one file
                • Brian Madison's template-story.md does already include a task checklist as part of the story
                  • keeping the stage checklist with the story may help LLMs not get out of sync. I've noticed that sometimes it forgets to update the checklist.
                  • keeping the story as its own file may help reduce the context's that's needed. It's not always necessary to put the entire history of the current development effort into context.
                • I like the idea of using Epic vs Story, which may have more precedent in the LLM training data for planning than Phase and Stage, which were ideas invented for the purpose of describing how to do this to an LLM
              • I also like the idea of having both epics and stories, which aren't quite the same as stages and todos. stories already include groups of todos, so there's really a three level vs two level hierarchy here. That seems useful.
              • One thing I still like about my workflow vs the one described here is that there is a ticket scoping mechanism, so two people could potentially work on items at the same time. It's not as "Single Player" as Harper Reed says:

                • • LLM coding is largely Single-player

                    • My main complaint about these workflows is that it is largely a solo endeavor - i.e. the interfaces are all single player mode.
                    • I have spent years coding by myself, years coding as a pair, and years coding in a team. It is always better with people. These workflows are not easy to use as a team. The bots collide, the merges are horrific, the context complicated.
                    • I really want someone to solve this problem in a way that makes coding with an LLM a multiplayer game. Not a solo hacker experience. There is so much opportunity to fix this and make it amazing.
              • That said, I still think the "endgame" setup is to use MCP to store all this information directly in a ticketing system as per CursorAI/Idea/Proj Rules GitHub CLI AI SLDC/MCP/Github Action Reasoning Model Planner
          • The project structure also includes an xnotes directory which is intended to be used for drafting longer prompts to copy andpaste into CursorAI/Agent. AFAICT this directory is only meant as a tempfile staging area for a human crafting prompts; it's not critical to read and can be ignored. Expand this node if you want more info.

            • this directory is added to CursorAI/.cursorindexingignore here and also added to .cursorignore here.
              • Side-note: I'm both relieved and dismayed to see others adding the same paths to both files. I've argued for better documentation on when people should use one vs the other in CursorAI/Forum/24/12/Questions on .gitignore, .cursorignore, .cursorban here.
            • This directory contains a few things for an Agile-like workflow with project rules.
              • xnotes/801-workflow-agile.mdc is a cursor project rule which describes the relationship between files used in coding. AFAICT this isn't actually in use, because it's not in the right directory; perhaps it's a draft.
              • xnotes/workflow-agile.md looks like another take at the agile workflow. Again, this isn't in the proper cursor rules directory, so the only way to reference it is to at message it.

• GraphRAG is famous for being expensive
  • alternative is LightRAG/Paper [2410.05779] LightRAG: Simple and Fast Retrieval-Augmented Generation
    • instead of 600 calls it just needs one
    • just testing them, it's better than naive rag
    • did it with AI/Model/Quen/2.5/7B quen2.5 7b
• compares it to OLAP

LangChain/GitHub/open_deep_research

• • Official Summary

    • Available search tools:
      • Tavily API - General web search
      • Perplexity API - General web search
      • Exa API - Powerful neural search for web content
      • ArXiv - Academic papers in physics, mathematics, computer science, and more
      • PubMed - Biomedical literature from MEDLINE, life science journals, and online books
      • Linkup API - General web search
      • DuckDuckGo API - General web search
      • Google Search API/Scrapper - Create custom search engine here and get API key here
    • Open Deep Research uses a planner LLM for report planning and a writer LLM for report writing:
      • You can select any model that is integrated with the init_chat_model() API
      • See full list of supported integrations here

• mise/GitHub/Issue/24/02/Struggles with integrating 1Password op
  • references mise/GitHub/Discussion/24/01/Vaults and Secrets Management
    • Tools discussed:
      • Secrets Manager
        • 1Password/Dev/op - 1Password CLI
        • sops - Mozilla's secret management tool
        • rops - Rust-based alternative for AWS/KMS
        • HashiCorp/Vault - Enterprise secret management
        • dotenv-org/vault - Encrypted .env files
        • keybase - Personal secret management
        • AWS/SSM/Param Store - AWS Systems Manager Parameter Store
      • Integration Tools:
        • direnv - Directory-based env management
        • Person/Vincent Prouillet/GitHub/tera - Template library used by mise
        • mise/plugin/mise-sops - sops plugin for mise

• 1Password/Blog/24/12/Replace .env with 1Password SDKs

  • • My Notes

      • how to migrate from EnvVar/.env files to 1Password's secure tooling
      • 1 - how to use 1Password/Dev/op/run and a 1Password Shared Vault to share secrets with other members on the dev team and inject them in to the EnvVar/.env file
      • 2 - how a remotely running automated script can use a 1Password Service Account and the 1Password/Dev/SDK/python to access secrets
      • comments - after this article, I have a better sense of how 1Password aims to cover the whole dev lifecycle.
        • I have an active #Question about the trade-offs of using this vs or with cloud-focused tools like HashiCorp/Vault or AWS/Secrets Manager. Duplication of secrets isn't fun, but convincing #DevOps to to use a tool that's convenient for development instead of HashiCorp Vault or an AWS-specific tool sounds even less fun. Maybe someone has made an integration ... ?
        • This article doesn't mention 1Password/Dev/Connect Server, which is the next level up from using a 1Password Service Account, and offers low latency and self-managed rate limits at the expense of self-hosting, etc. See 1Password/Dev/Doc/secrets-automation/Overview/Comparison
  • VSCode/Extension/1Password
    • • Select a secret in EnvVar/.env file and use VSCode/Extension/1Password/Command/Save in 1Password VSCode/Command to swap it for a secret reference
        • Before
          • bash

            # This is an example .env file that we will use to migrate  
            # from hosting the actual configurations of an environment  
            # over to a 1Password-managed environment. 
            
            # Environment Configuration  
            # options include  
            #         - development  
            #         - staging  
            #         - production  
            APP_ENV=development
            
            # Application Configuration  
            APP_NAME=OP SDJ ENV Demo Application
            
            APP_SECRET_TOKEN=QAZWSXEDCrfvtgbyhn 

        • highlight line 15, then use VSCode Save in 1Password
        • After, it's replaced with 1Password/Dev/CLI/Secret Reference
          • first, they move it to a generic vault named "Vault"
            • 
  • • Example Python remote script using 1Password Service Account
      • python

        import asyncio
        import os
        from onepassword import Client
        
        async def main():
          # You still need to load your Service Account Token from the OS, 
          # but let's do it without an additional package.  
          token = os.getenv("OP_SERVICE_ACCOUNT_TOKEN")
          client = await Client.authenticate(auth=token, 
                                           integration_name="DevRel Demo",
                                           integration_version="v0.0.1")
          
          app_name = await client.secrets.resolve("op://ENV_Demo_Secrets/Demo App Name/text")
          app_secret_token = await client.secrets.resolve("op://ENV_Demo_Secrets/Blog 1 ENV App_Secret_Token/Section_ty4kl2xveagt5wxcz4yzfzloia/token")

      • My Note If we've instrumented this so it can run as a remote job and use a 1Password Service Account, I think the assumption that locally, I'll then use 1Password/Dev/op op to get a 1Password/Dev/CLI/Secret Reference to load the OP_SERVICE_ACCOUNT_TOKEN ... is that right?
• 1Password/Dev/Doc/sdks

  • • My Notes

      • requires a 1Password Service Account
• 1Password/Dev/Connect Server
  • 1Password connect diagram

  • • Comparison table of when to use 1Password Service Account vs 1Password/Dev/Connect Server

      • Feature	Service accounts	Connect servers
        1Password CLI	✅	✅
        REST API	❌	✅
        Self-hosted	❌	✅
        Infrastructure requirements	❌	✅
        Low overhead	✅	❌
        Low latency	❌	✅
        Kubernetes integrations	✅	✅
        CI/CD integrations	✅	✅
        Rate limits	✅	❌
        Request quotas	✅	❌
        Included with 1Password subscription	✅	✅

• Person/Eugene Wei/Blog/19/02/Status as a Service - StaaS

1Password/Dev/How To/Replace .env secrets with op refs

• 1Password/How To/Move item from one account to another
• 1Password/Dev/Doc/sdks
• 1Password/Dev/SDK/python
• 1Password Service Accounts
• mise/GitHub/Issue/24/02/Struggles with integrating 1Password op
  • CLI/Tool
    • #Discovered direnv

• Logseq/Slideshow/How To/Create a Logseq Slideshow
• Logseq/Doc/slide
• #Discovered Person/Richard Yu/GitHub/logseq-plugin-slide-reveal
  • This is a paid plugin
• Logseq/Publish/How To/Publish with a Theme