Add AI-OS security research layer#8
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 8bcc82534f
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
|
||
| ## AI-Assisted Security Research OS | ||
|
|
||
| This repo also includes an [AI-OS](ai-os/README.md) layer for using AI assistants responsibly during authorized bug hunting. The AI-OS defines operating principles, safety gates, modes, prompts, rubrics, schemas, and synthetic examples for AI-assisted evidence review and report preparation. |
There was a problem hiding this comment.
AGENTS.md requires knowledge/log.md to be updated when adding major documentation. This commit introduces the new AI-OS documentation layer and links it from the README, but knowledge/log.md still only records the 2026-05-20 baseline, so the repository log omits this major addition.
Useful? React with 👍 / 👎.
| if (line.toLowerCase().includes(phrase) && !educationalPhraseFiles.has(rel)) { | ||
| warnings.push(`${rel}:${lineNo}: review unsafe phrase '${phrase}'`); |
There was a problem hiding this comment.
In npm run check:ai-os, non-exempt AI-OS files containing phrases such as bypass rate limits are only added to warnings, and the script still exits successfully because only failures are fatal. That means a new unsafe recommendation in an AI-OS Markdown file can pass CI with just a console warning, weakening the safety boundary this check is meant to enforce.
Useful? React with 👍 / 👎.
1 participant
Summary
Validation