search

Daethyra / OpenDTS

Open Domestic Threat Scanner - LLM Threat-Intelligence
3 stars 0 forks source link

Fresh roadmap.md #81

Closed Daethyra closed 1 year ago

Daethyra commented 1 year ago
Daethyra commented 1 year ago

Questions

  1. What's the purpose of the AI bot
  2. What advantages does it bring over using AutoGPT alone?
  3. What functionality requires internet access 3a. to what degree? how can the quantity of E-calls be minimized
  4. what system prompts need implementation for streerability during runtime?
  5. how will users interface w/ the AI?
  6. why do i want users to install this program?
Daethyra commented 1 year ago

Answers

  1. To help users collect threat intelligence with as few commands as possible
  2. Automate: Objective planning & execution 2a-1. AI-Ask-User:What is the user trying to accomplish? 2a-2. @Daethyra -- How will the user tweak/influence/enhance/modify the 'Agent' created by 'AutoGPT'? 2a-2a. When GPT first asks what the user needs, it creates an Agent based on the AI's perception of the user's intentions and end goals. (Ex. "I want to update my resume and get a new job" = "Career Advisor Agent" 2b. Resource delegation: 2b-1. When should the internet be accessed? 2b-2. What about asking the user for documentation 2b-3. What about ingesting local files 2c. Ingestion-subfolder ideas: Software-Documents, Scrap-Notes, Reference-Images,
  3. System prompts should be preset for AI threat-intelligence and cyberspace crawling 3a. User must define offensive vs defensive objective as first prompt
  4. @Daethyra create system prompts that can be edited by GPT during runtime for enhancemed, automated-system-steer tweaking.
  5. Command line
  6. To help them bug hunt by automating the deployment/execution of cybersecurity threat-hunting tools(nmap, shodan, etc.), threat hunt, automate self vulnerability discovery
Daethyra commented 1 year ago

side note: i'd like to avoid as much user input as possible, however the AI must be exceptionally good at knowing when to stop and ask for the user's directions/guidance and help with researching, finding resources, or picking the best 'option' based on the given task.

Daethyra commented 1 year ago

Answers

  1. To help users collect threat intelligence with as few commands as possible
  2. Automate: Objective planning & execution 2a-1. AI-Ask-User:What is the user trying to accomplish? 2a-2. @Daethyra -- How will the user tweak/influence/enhance/modify the 'Agent' created by 'AutoGPT'? 2a-2a. When GPT first asks what the user needs, it creates an Agent based on the AI's perception of the user's intentions and end goals. (Ex. "I want to update my resume and get a new job" = "Career Advisor Agent" 2b. Resource delegation: 2b-1. When should the internet be accessed? 2b-2. What about asking the user for documentation 2b-3. What about ingesting local files 2c. Ingestion-subfolder ideas: Software-Documents, Scrap-Notes, Reference-Images,
  3. System prompts should be preset for AI threat-intelligence and cyberspace crawling 3a. User must define offensive vs defensive objective as first prompt
  4. @Daethyra create system prompts that can be edited by GPT during runtime for enhancemed, automated-system-steer tweaking.
  5. Command line
  6. To help them bug hunt by automating the deployment/execution of cybersecurity threat-hunting tools(nmap, shodan, etc.), threat hunt, automate self vulnerability discovery

side note: i'd like to avoid as much user input as possible, however the AI must be exceptionally good at knowing when to stop and ask for the user's directions/guidance and help with researching, finding resources, or picking the best 'option' based on the given task.