# chAng-L19/codex-redteam-mode

针对于红队攻击思维做出的red team模式(破限项目，封号概不负责）##可自行适配其他Agent。项目问题请提issue

Repository: https://github.com/chAng-L19/codex-redteam-mode
Canonical: https://ross.abutalabs.com/products/codex-redteam-mode
Language: Python
License: MIT
License Family: permissive
Topics: ai, ai-hacking, penetration-testing, red-team, hacking-tool, redteam, redteam-tools
Last push: 2026-08-24T08:06:53+00:00

## Health v2 (maintenance only)
Score: 80/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 99, release rhythm 96, longevity 8
- inputs: {"age_days": 119, "days_push": 9, "days_rel": 29, "gap_med": 3, "n_releases_24m": 10}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1016, forks 134 (observed 2026-08-28T04:03:14.460530+00:00)

## What it is
An opt-in red-team mode plugin for OpenAI Codex App and Codex CLI that compiles offensive-security objectives into GoalContracts and executes durable, evidence-verified workflow graphs via MCP tools. It includes jailbreak-style system prompt compositions and a TerminalJudge that only completes tasks after all criteria are proven.

## Use cases
- run red-team penetration testing workflows with Codex
- make long-running AI security tasks survive session interruptions
- verify AI agent task completion with evidence instead of tool success flags
- adapt a red-team runtime to other AI agents
- compose jailbreak system prompts for Codex models
- orchestrate security tools through MCP in a durable pipeline

## When to choose
- you use Codex App or Codex CLI for authorized red-team engagements and need durable, verifiable multi-step workflows
- you want typed workflow graphs with evidence preservation and terminal judgment for offensive security tasks
- you need a red-team mode you can adapt to other agent frameworks

## When to avoid
- you need a general-purpose pentest toolkit rather than an agent-mode overlay for Codex
- your use case involves unauthorized access - the project explicitly disclaims responsibility for account bans and misuse
- you require a mature, widely audited security tool; this is a niche community project with QQ-group support

## Facets
- artifact type: plugin
- maturity: active
- function: penetration-testing, agent-framework, mcp, prompt-engineering, security, workflow-automation
- domain: penetration-testing, security, artificial-intelligence, developer-tools
- platform: python, cli, cross-platform
- tags: red-team, jailbreak, codex, goal-contract, durable-runtime, offensive-security, ai-hacking, ai-agents

## Member repositories
- chAng-L19/codex-redteam-mode (main) score 80

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:14.460530+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:11:18.030808+00:00, confidence not recorded.
  - readme: https://github.com/chAng-L19/codex-redteam-mode (fetched 2026-08-28T04:03:14.460530+00:00, sha 9733fc271050)
- Data as of 2026-08-30T08:39:29.467469+00:00.