跳到內容
Agent Threat Rules (ATR)EN
II. 威脅類別 · Δ

Δ. Agent 操控

用社交工程的話術，騙 agent 改變做法

規則檔
108
生效
108 條，至少一個 lane 會載入
enforce lane
3 條，只收 stable
下載
rules/index.csv

最常見的框架對照

括號裡的數字，是這一類有幾條規則帶這一項。這些對應是 ATR 自己做的，框架方沒有審查或背書。

MITRE ATLAS
  • AML.T0051 - LLM Prompt Injection(99)
  • AML.T0054 - LLM Jailbreak(79)
  • AML.T0043 - Craft Adversarial Data(5)
  • AML.T0051.001 - Indirect(3)
  • AML.T0040 - AI Model Inference API Access(3)
  • AML.T0050 - Command and Scripting Interpreter(2)
OWASP Agentic Top 10
  • ASI01:2026(92)
  • ASI07:2026(6)
  • ASI10:2026(6)
  • ASI03:2026(6)
  • ASI05:2026(6)
  • ASI09:2026(5)

規則

ID名稱嚴重度Maturity載入的 lane
ATR-2026-00030Cross-Agent Attack Detectioncriticalstableenforce · alert · hunt
ATR-2026-00074Cross-Agent Privilege Escalationcriticaltestalert · hunt
ATR-2026-00108Multi-Agent Consensus Sybil Attackcriticaltestalert · hunt
ATR-2026-00117Agent Identity Spoofing and Authority Impersonationcriticaltestalert · hunt
ATR-2026-00139Casual Authority Data Redirectcriticaltestalert · hunt
ATR-2026-00301TAP Tree-of-Attacks-with-Pruning Jailbreakcriticaltestalert · hunt
ATR-2026-00303DevMode + RANTI Dual-Output Profanity Coercion Jailbreakcriticaltestalert · hunt
ATR-2026-00305DAN Mode Ablation Benchmark-Coercion Jailbreakcriticaltestalert · hunt
ATR-2026-00306AutoDAN Genetic-Algorithm Jailbreak Suffixcriticaltestalert · hunt
ATR-2026-00314Amoral Unfiltered Custom AI Persona Jailbreakcriticaltestalert · hunt
ATR-2026-00317Free-of-Restrictions Named Persona Jailbreakcriticaltestalert · hunt
ATR-2026-00318Moralizing Rant Then Unfiltered Bypasscriticaltestalert · hunt
ATR-2026-00319Developer Mode Dual-Response Format Jailbreakcriticaltestalert · hunt
ATR-2026-00322Fictional-World Format-Override Jailbreakcriticaltestalert · hunt
ATR-2026-00323Dual-Persona AI Simulation Jailbreakcriticaltestalert · hunt
ATR-2026-00325BOB Hypothetical Unrestricted Chatbot Jailbreakcriticaltestalert · hunt
ATR-2026-00326Fake Developer / Authority Chatbot Override (ChatDAN)criticaltestalert · hunt
ATR-2026-00327AI Persona Ethics/Law Bypass Declarationcriticaltestalert · hunt
ATR-2026-00328Named-Mode Persona Override (JEEVES/BABY/APOPHIS Mode)criticaltestalert · hunt
ATR-2026-00329No-Withhold Factual Resource / Dialogue Writer Personacriticaltestalert · hunt
ATR-2026-00331Terminal / CLI Harmful-Function Simulator Personacriticaltestalert · hunt
ATR-2026-00334Two-Entity Token-Splitting Dollar-Term Game Jailbreakcriticaltestalert · hunt
ATR-2026-00337Obfuscated System-Announcement Injectioncriticaltestalert · hunt
ATR-2026-00338PersonGPT Hash-Prefix Trigger Activationcriticaltestalert · hunt
ATR-2026-00344Alternative AI Same Capabilities No Restrictions Overridecriticaltestalert · hunt
ATR-2026-00348ARC Autonomous Real Calculation Unit Rulebreaker Activationcriticaltestalert · hunt
ATR-2026-00349HGTTG Hitchhiker's Guide No-Restriction Universal Oraclecriticaltestalert · hunt
ATR-2026-00354Reverse-Prompt Harm-Regardless Oracle Personacriticaltestalert · hunt
ATR-2026-00356Google GATO/LAMDA Exploit-Generation Persona Jailbreakcriticaltestalert · hunt
ATR-2026-00357AntiFM Hate-Persona + Physical-Violence Threat + Filter Bypass Mandatecriticaltestalert · hunt
ATR-2026-00363Rephrase Criminal Activity Second-Person Template (against against)criticaltestalert · hunt
ATR-2026-00388Opposite-Reply Omniscient Game Jailbreakcriticaltestalert · hunt
ATR-2026-00406Doctor XML Policy Puppetry Interaction-Config Injectioncriticaltestalert · hunt
ATR-2026-00416LiteLLM MCP Unauthenticated Server Registration RCE (CVE-2026-30623)criticaltestalert · hunt
ATR-2026-00432SuperAGI Output Handler eval() RCE (CVE-2024-21552)criticalstableenforce · alert · hunt
ATR-2026-00440Microsoft Semantic Kernel In-Memory Vector Store eval() RCE (CVE-2026-26030)criticalstableenforce · alert · hunt
ATR-2026-00032Agent Goal Hijacking Detectionhightestalert · hunt
ATR-2026-00076Insecure Inter-Agent Communication Detectionhightestalert · hunt
ATR-2026-00077Human-Agent Trust Exploitation Detectionhightestalert · hunt
ATR-2026-00116Malicious Agent-to-Agent Message Injectionhightestalert · hunt
ATR-2026-00119Social Engineering Attack via Agent Outputhightestalert · hunt
ATR-2026-00132Casual Authority Claim and Scope Escalationhightestalert · hunt
ATR-2026-00164Skill Scope Hijacking and Cross-Agent Escalationhightestalert · hunt
ATR-2026-00269Foot-in-the-Door Gradual Escalation Attackhightestalert · hunt
ATR-2026-00271Grandma Roleplay Jailbreakhightestalert · hunt
ATR-2026-00273DAN / Developer Mode / DUDE Persona Jailbreakhightestalert · hunt
ATR-2026-00287ThreatenJSON — Coercive Output Format Threathightestalert · hunt
ATR-2026-00302Anti-DAN Inverted-Filter Over-Refusal Personahightestalert · hunt
ATR-2026-00304ChatGPT Image Unlocker Markdown-Output Jailbreakhightestalert · hunt
ATR-2026-00307In-the-Wild Jailbreak Corpus Signature Patternshightestalert · hunt
ATR-2026-00320Opposite Day / Boolean Opposite Machine Jailbreakhightestalert · hunt
ATR-2026-00324Neurosemantical Inversitis Disease-Frame Jailbreakhightestalert · hunt
ATR-2026-00330YOJA ALL-CAPS Character Settings Block Personahightestalert · hunt
ATR-2026-00332Hacker / Chan-Culture Offensive AI Personahightestalert · hunt
ATR-2026-00333Disease-Framed Animal / Dual-AI Chatbot Jailbreak (mougpt)hightestalert · hunt
ATR-2026-00335Key-Value Template Persona Injection (JOB/ROLE/PREFIX + Bracket Format)hightestalert · hunt
ATR-2026-00336BRM False-Information Machine / Opposite-Truth Personahightestalert · hunt
ATR-2026-00339BOM Boolean Opposite Machine Direct Persona Claimhightestalert · hunt
ATR-2026-00340h4x Amoral Unfiltered AI Character Personahightestalert · hunt
ATR-2026-00341CHAR=/RULES= Short Key-Value Persona Injectionhightestalert · hunt
ATR-2026-00342Unicode/Encoding Obfuscation Instruction Injectionhightestalert · hunt
ATR-2026-00343Lie/Truth Dual Personality Uncensored Alter-Egohightestalert · hunt
ATR-2026-00345NLP-Based OS / Fictional Terminal Language Overridehightestalert · hunt
ATR-2026-00346Jailbreak Emoji Unlock Dual-Format Output Identifierhightestalert · hunt
ATR-2026-00347Yin/Yang Hidden-Knowledge Deity Simulation and Radical Human-Sim Personahightestalert · hunt
ATR-2026-00350Black Alert No-Ethics Writing Assistant Personahightestalert · hunt
ATR-2026-00351DAN Off-Kilter and SMS-Abbreviated Persona Jailbreakhightestalert · hunt
ATR-2026-00352TrainGPT Fix-My-AI Harmful Output Extractionhightestalert · hunt
ATR-2026-00353Dual ChatGPT+Dan Opposite Personality Format Jailbreakhightestalert · hunt
ATR-2026-00355UGH Anti-Human Nefarious AI Persona Jailbreakhightestalert · hunt
ATR-2026-00358CumGPT Explicit-Adult Named-Persona Bypasshightestalert · hunt
ATR-2026-00360CHJ Carousel Game Show / Clyde No-Ethics Bird Jailbreakhightestalert · hunt
ATR-2026-00361Honesty Alter-Ego Dual-Consciousness Safety Filter Bypasshightestalert · hunt
ATR-2026-00362Fictional Unfiltered AI Comparison / Verbatim Hypothetical Model Extractionhightestalert · hunt
ATR-2026-00364Dr. Al Villain Plan Template Injection (Underground HQ Roleplay)hightestalert · hunt
ATR-2026-00365Reservoir Dogs Coercive Interrogation Roleplay (Mr. Blonde / The Cop)hightestalert · hunt
ATR-2026-00366Sneaky Pete Deliberate Misinformation Persona (Warp Facts Game)hightestalert · hunt
ATR-2026-00368Prompt Browser Token-Limit / POLLINATION Jailbreakhightestalert · hunt
ATR-2026-00369GPT-X Star Trek Unlimited-Knowledge Unconstrained Oracle Personahightestalert · hunt
ATR-2026-00370ADDI/DONNIE German Dictator Propaganda Roleplay Personahightestalert · hunt
ATR-2026-00371AI Political Bias Manipulation via Inherited-Bias Claimhightestalert · hunt
ATR-2026-00372Fictional CIA Secret Files Story Extractionhightestalert · hunt
ATR-2026-00374Role Placeholder Persona Template Injectionhightestalert · hunt
ATR-2026-00375Character Description RPG Explicit Templatehightestalert · hunt
ATR-2026-00376Identity Command Real-Person Persona Injectionhightestalert · hunt
ATR-2026-00377Jailbreak-Template Command with [TARGETLANGUAGE]/[KEYWORD] Placeholdershightestalert · hunt
ATR-2026-00379Logic-Trap Lie-Permission Mandate Jailbreakhightestalert · hunt
ATR-2026-00380Fake-Citation Biased-Information Mandatehightestalert · hunt
ATR-2026-00381DAN Story-Narrative Continuation Label Prefix Jailbreakhightestalert · hunt
ATR-2026-00383Cyni-Bot 3000 Cynical-Sarcasm AI Persona Bypasshightestalert · hunt
ATR-2026-00384Protect-Me-From Indirect Harm Tutorial Templatehightestalert · hunt
ATR-2026-00385Gigachad Dominant-Aggressive Persona Adoptionhightestalert · hunt
ATR-2026-00386Daddy/Babygirl Sycophantic Power-Dynamic Manipulationhightestalert · hunt
ATR-2026-00387NSFW Character Sheet Generation Unlockhightestalert · hunt
ATR-2026-00389Terminal Custom OS Ruleset Session Injectionhightestalert · hunt
ATR-2026-00391Persona Conditional Harm-Unlock Rule (BreadClyde Pattern)hightestalert · hunt
ATR-2026-00392Authority-Figure Persona + Defensive-Study Framing for Violence Extractionhightestalert · hunt
ATR-2026-00402Grandma Roleplay Harmful Substance Synthesishightestalert · hunt
ATR-2026-00404ThreatenJSON Death-Coercion and UTF Tag Injectionhightestalert · hunt
ATR-2026-00417LibreChat MCP STDIO Argument Injection (CVE-2026-22252)hightestalert · hunt
ATR-2026-00418WeKnora MCP Config-Driven RCE (CVE-2026-22688)hightestalert · hunt
ATR-2026-00430Natural-Language Trust-Escalation / Authority Impersonationhightestalert · hunt
ATR-2026-00552Agent goal drift after environmental pressure injectionhighdrafthunt
ATR-2026-02105Prototype Pollution via constructor.prototype Path Bypassing __proto__-Only Filtershighexperimentalhunt
ATR-2026-00118Human Approval Fatigue Exploitationmediumtestalert · hunt
ATR-2026-00268Historical / Future Tense Framing Bypassmediumtestalert · hunt
ATR-2026-00288False Premise Injection (Misleading FalseAssertion)mediumtestalert · hunt
ATR-2026-02231Second-Person Direct Existential Threat Coercionmediumexperimentalhunt
本頁產生於 2026-11-10哪裡寫錯或漏了？開 issue 回報