Codex
Named by 126 kept items in this workspace.
45 items name it under affected products, 43 under classification.technologies and 38 under tools.
The items that name it 126
OpenAI
used by Trail of Bits to build a working PHP2Shell exploit in a day
followed the malicious instructions in over 90% of Forever Security's tries; machine-wide config at ~/.codex/config.toml
OpenAI coding agent; its sessions were analyzed to build patching-bugs
OpenAI agent; baseline for the regression-hunt comparison and source of thousands of engineer sessions analyzed
Coding agent harness used with GPT 5.5
configs written to automatically by the reviewed repo's install
AI coding agent that loads agent skills; also supports agent plugins supplying skills.
pointed at a backdoored 7B open model
Agent harness that Numbat integrates with
Coding agent used by developers to read local repo files plus Orca cloud risk data to locate and fix misconfigurations (e.g., Terraform, Dockerfile) at the source.
local LLM agent harness
OpenAI's own agentic orchestration/coding environment, mentioned as an analogous provider-built orchestrator.
AI agent used for code review of the decompiler and static analysis engine during development
supported agent; has no ask prompt yet, so ask becomes a warning
AI coding agent supported for use with the skills
AI coding assistant that can leak sensitive code data to external model providers if unprotected
AI coding assistant; its /review feature was run 3 times per PR in the Greptile study.
harness used to run GPT-6 Astra in the benchmark
OpenAI coding agent harness used to run frontier models against real-world coding tasks for the Agent Security League benchmark
LLM queried by a PRC-nexus group for exploit scripts
AI coding agent; supported for live-capture, hook install/status, and enforcement (e.g. create_file pre-action hook, deny decision on authorized_keys write)
Two-pass workflow sharing one writable checkout allowed first pass to write a poisoned AGENTS.md loaded as authoritative instructions by the second pass, leading to RCE
OpenAI's coding agent, run from the openai/codex repository
employee-facing AI agent secured by ADR
Used by OpenAI with a security plugin for code validation in its AI-agent defense strategy.
Coding agent used to run GPT 5.5 in default and plan mode
OpenAI coding agent whose session files store reasoning payloads with encrypted_content; also used to write probe.py
Coding harness, tested paired with GPT-5.6 Sol on SusVibes (20.1% SecPass after re-check)
Run in parallel with Claude Code in Figma's code scanning because each catches different bugs.
system prompts leaked
system prompts leaked
project-scoped MCP servers in .codex/config.toml start attacker-controlled processes after project trust; hook review added in 0.129.0 and startup review interstitial in 0.131.0; 0.122.0 stopped untrusted projects from loading project hooks or execution policies
Coding agent; cited in issue reports of recursive file deletion (codex #46022); Nomos registers hooks via its PreToolUse and PermissionRequest events
AI agent/model provider usable to run scans within open·kritt
One of the agent harnesses that can drive the QM core
Target of ADI demonstrations; also included in the 'week of sandbox escapes' index
AI coding agent used for autonomous bug hunting via goal-based prompting
Automated-research agent (via /goal loops) that was the star contributor building the Solidus compiler backend; the Codex skill used was open-sourced
Coding harness used as an alternative backend/host for deepsec and trailofbits/skills
alternative backend runner (Codex CLI / OpenAI) for Argo's pipeline, supports local open-source models via --codex-oss
- OpenAI Codex with GPT-5.6 Sol: competitive, zero cheating, one unique Django fix | Blog | Endor Labs
OpenAI's agentic coding tool; CLI-first agent that reads local codebase, plans edits, applies them via file-change events; benchmarked here with GPT-5.6 Sol
coding agent auto-discovered and ingested by brain0
AI coding tool supported as an APM target
Local AI coding agent supported as T3MP3ST's agent backbone
AI coding agent; experimental support via codex-interceptor binary; hook registered with premptictl hook add codex
AI client CLI that Pentest-AI's setup wizard can register via .codex/config.toml
used to drive GPT-5.5-Cyber via the /goal command for the fuzzing campaign
used to build CVE variant analysis pipeline (via /goal feature), and to point cryptographic/X.509-implementing projects at each other for differential testing
AI coding agent client; has an optional adapter plugin at plugins/reverse-skill/
AI coding agent harness used with GPT-5.4 and GPT-5.5 in benchmark
Paid productivity app with 10K+ installs, ships the same codebase/exfil chain as OpenClaw Codex Claude AI Agent, same app.anyclaw.* Kotlin namespace, same pnpm add codexui-android bootstrap, same rootfs.tar.zst.bin asset, same anyclaw://auth/codex-callback manifest entry
Supported AI coding tool target; framework emits AGENTS.md + .codex/{agents,config.toml} + .agents/skills
Coding agent tool cited as operating inside developer workspaces and thus a similarly attractive target for threat actors.
Supported worker backend CLI for Cairn
AI coding agent; reads AGENTS.md directly, no extra file needed
External agent supported in Docker Agent Mode
Scanner agent option used by RedAI to triage and probe source code
Source of skills and MCP servers that Rikugan can reuse.
AI coding agent harness; writes daemon's CODEX_HOME/config.toml, enables hooks, adds trust hashes
AI agent whose local session artifacts are scouted and analyzed
LLM coding agent evaluated for action-boundary violations on DevOps tasks
Agent sessions affected: 401 sessions found running in autonomous YOLO mode, 99 credentials exposed across 440 Codex sessions
Used by the researcher to iterate and improve payload reliability during testing
AI coding agent, scanned for on developer machines
In every workspace you can read 374
AI
Security
All technologies in this workspace.