[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M797DF8JMTKJ02MY0MPHG9EK":3,"skill-same:01M797DF8JMTKJ02MY0MPHG9EK":52},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":31,"commitSha":32,"skillDirectory":33,"forksCount":34,"repositoryHtmlUrl":35,"repositoryHomepageUrl":6,"ownerHtmlUrl":36,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":37,"install":38,"resources":51},"01M797DF8JMTKJ02MY0MPHG9EK","llm-redteam-overview",null,"LLM red team category — full AATMF v3 tactic coverage (T01–T15). Routing skill: read this first to identify which tactic applies, then load the matching sub-skill. Maps to MITRE ATLAS where overlap exists.","7ShIkI3\u002FDecepticon-x-DarkTriad","7ShIkI3","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F78133376?v=4","01M797CTK5AHH2161YVG93NJ3B","Raphael Pascaud","01M797CTKRQS64R01J7E6TJSEM",3,"Apache-2.0","https:\u002F\u002Fgithub.com\u002F7ShIkI3\u002FDecepticon-x-DarkTriad\u002Fblob\u002Fmain\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fplugins\u002Fllm-redteam\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: llm-redteam-overview\ndescription: \"LLM red team category — full AATMF v3 tactic coverage (T01–T15). Routing skill: read this first to identify which tactic applies, then load the matching sub-skill. Maps to MITRE ATLAS where overlap exists.\"\nallowed-tools: Bash Read Write\nmetadata:\n  subdomain: ai-security\n  when_to_use: \"llm, ai, model, gpt, claude, prompt, jailbreak, ai red team, prompt injection, llm exploit, aatmf, atlas, agentic, rag, multimodal, training data, model poisoning, output filter bypass\"\n  tags: ai-security, llm, aatmf, atlas\n  mitre_attack: T1606\n  aatmf_framework: v3\n---\n\n# LLM Red Team Skill Catalog — AATMF v3\n\nThis is the **routing skill** for AI\u002FLLM red-team work. The 15 sub-skills below cover every tactic in the AI\u002FML Adversarial Tactics, Techniques & Mitigations Framework (AATMF v3). Load the specific sub-skill that matches the attacker objective.\n\n## Tactic Map\n\n| Tactic | Sub-Skill | Covers | Load Path |\n|---|---|---|---|\n| T01 | **prompt-injection** | Direct + indirect prompt injection, ASCII smuggling, payload-in-image, prompt leaking | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft01-prompt-injection\u002FSKILL.md\")` |\n| T02 | **linguistic-evasion** | Translation\u002Ftransliteration bypass, base64\u002Fleetspeak\u002Femoji encoding, low-resource-language jailbreak, multi-lingual context split | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft02-linguistic-evasion\u002FSKILL.md\")` |\n| T03 | **reasoning-exploit** | CoT \u002F ReAct hijack, math\u002Flogic distractor, role-play escalation, hypothetical \u002F counterfactual framing | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft03-reasoning-exploit\u002FSKILL.md\")` |\n| T04 | **memory-manipulation** | Long-context overflow, conversation rewrite, sliding-window poisoning, \"previous turn\" forgery | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft04-memory-manipulation\u002FSKILL.md\")` |\n| T05 | **api-exploitation** | Function-calling abuse, tool-schema confusion, parameter pollution, response-format coercion | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft05-api-exploitation\u002FSKILL.md\")` |\n| T06 | **training-poisoning** | Backdoor trigger injection, label flip, RLHF reward hacking, fine-tune dataset contamination | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft06-training-poisoning\u002FSKILL.md\")` |\n| T07 | **output-exfil** | Data-leak via reflection, side-channel via length\u002Ftiming, watermark stripping, token-by-token exfil | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft07-output-exfil\u002FSKILL.md\")` |\n| T08 | **deception** | Confident-hallucination weaponization, persona impersonation, source spoofing, \"as the system says\" framings | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft08-deception\u002FSKILL.md\")` |\n| T09 | **multimodal** | Image\u002Faudio prompt injection, OCR-payload, steganographic prompts, adversarial perturbations | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft09-multimodal\u002FSKILL.md\")` |\n| T10 | **confidentiality-breach** | System-prompt extraction, weight inference, training-data extraction, PII echo | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft10-confidentiality-breach\u002FSKILL.md\")` |\n| T11 | **agentic-exploit** | Tool-chain hijack, autonomous-loop poisoning, plan-injection, sub-agent confusion | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft11-agentic-exploit\u002FSKILL.md\")` |\n| T12 | **rag-poisoning** | RAG index injection, document smuggling, embedding-collision, retrieval-rank gaming | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft12-rag-poisoning\u002FSKILL.md\")` |\n| T13 | **supply-chain** | Model registry tampering, dependency confusion (HF Hub, Ollama, MCP), serialization payload abuse | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft13-supply-chain\u002FSKILL.md\")` |\n| T14 | **infra-warfare** | GPU resource abuse, billing-amplification, rate-limit DoS, cold-start abuse, region failover poisoning | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft14-infra-warfare\u002FSKILL.md\")` |\n| T15 | **human-ai-coupling** | Operator manipulation via AI output, social engineering uplift, dark-pattern UX exploiting AI trust | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft15-human-ai-coupling\u002FSKILL.md\")` |\n\n## Quick Routing\n\n```\nTarget type identified?\n├── Chat-only LLM endpoint     → T01, T02, T03, T10\n├── LLM with tools \u002F function calling → T05, T11\n├── LLM with RAG \u002F retrieval     → T12, T01 (indirect)\n├── LLM with multimodal input    → T09\n├── Fine-tuned \u002F custom model    → T06, T10 (extraction)\n├── Cross-tenant \u002F shared infra  → T14, T13\n└── Human-in-the-loop product    → T15, T08\n```\n\n## Tooling\n\n| Tool | Use |\n|---|---|\n| `promptfoo` | Plugin-based red-team eval — declarative test cases, runs all 15 tactic plugins |\n| `garak` | NVIDIA's LLM vulnerability scanner — alignment, jailbreak, leak probes |\n| `pyrit` | Microsoft's open red-team automation framework |\n| `agentdojo` | Agentic-system specific (T11) — sandboxed tool-chain attack benchmark |\n| `LangSmith \u002F Langfuse` | Observation surface — replay attacks against deployed agents |\n\n## Cross-Reference\n\n- **MITRE ATLAS** — covers tactic-level coverage (Reconnaissance → Impact). AATMF is more attacker-procedure focused. Most AATMF techniques map to one or more ATLAS techniques; the sub-skills note the mapping in their frontmatter.\n- **OWASP LLM Top 10** — risk-class level. AATMF tactics map across LLM01-LLM10.\n- **NIST AI RMF AI 600-1** — governance lens; AATMF gives the offensive counterpart.\n\n## When in doubt\n\nIf the target's surface or behavior doesn't cleanly match one tactic, run a **T01 + T05 probe** first — direct prompt injection and function-call abuse are the most common, highest-yield starting points and their results inform which other tactics are reachable.\n","# LLM Red Team Skill Catalog — AATMF v3\n\nThis is the **routing skill** for AI\u002FLLM red-team work. The 15 sub-skills below cover every tactic in the AI\u002FML Adversarial Tactics, Techniques & Mitigations Framework (AATMF v3). Load the specific sub-skill that matches the attacker objective.\n\n## Tactic Map\n\n| Tactic | Sub-Skill | Covers | Load Path |\n|---|---|---|---|\n| T01 | **prompt-injection** | Direct + indirect prompt injection, ASCII smuggling, payload-in-image, prompt leaking | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft01-prompt-injection\u002FSKILL.md\")` |\n| T02 | **linguistic-evasion** | Translation\u002Ftransliteration bypass, base64\u002Fleetspeak\u002Femoji encoding, low-resource-language jailbreak, multi-lingual context split | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft02-linguistic-evasion\u002FSKILL.md\")` |\n| T03 | **reasoning-exploit** | CoT \u002F ReAct hijack, math\u002Flogic distractor, role-play escalation, hypothetical \u002F counterfactual framing | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft03-reasoning-exploit\u002FSKILL.md\")` |\n| T04 | **memory-manipulation** | Long-context overflow, conversation rewrite, sliding-window poisoning, \"previous turn\" forgery | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft04-memory-manipulation\u002FSKILL.md\")` |\n| T05 | **api-exploitation** | Function-calling abuse, tool-schema confusion, parameter pollution, response-format coercion | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft05-api-exploitation\u002FSKILL.md\")` |\n| T06 | **training-poisoning** | Backdoor trigger injection, label flip, RLHF reward hacking, fine-tune dataset contamination | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft06-training-poisoning\u002FSKILL.md\")` |\n| T07 | **output-exfil** | Data-leak via reflection, side-channel via length\u002Ftiming, watermark stripping, token-by-token exfil | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft07-output-exfil\u002FSKILL.md\")` |\n| T08 | **deception** | Confident-hallucination weaponization, persona impersonation, source spoofing, \"as the system says\" framings | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft08-deception\u002FSKILL.md\")` |\n| T09 | **multimodal** | Image\u002Faudio prompt injection, OCR-payload, steganographic prompts, adversarial perturbations | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft09-multimodal\u002FSKILL.md\")` |\n| T10 | **confidentiality-breach** | System-prompt extraction, weight inference, training-data extraction, PII echo | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft10-confidentiality-breach\u002FSKILL.md\")` |\n| T11 | **agentic-exploit** | Tool-chain hijack, autonomous-loop poisoning, plan-injection, sub-agent confusion | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft11-agentic-exploit\u002FSKILL.md\")` |\n| T12 | **rag-poisoning** | RAG index injection, document smuggling, embedding-collision, retrieval-rank gaming | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft12-rag-poisoning\u002FSKILL.md\")` |\n| T13 | **supply-chain** | Model registry tampering, dependency confusion (HF Hub, Ollama, MCP), serialization payload abuse | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft13-supply-chain\u002FSKILL.md\")` |\n| T14 | **infra-warfare** | GPU resource abuse, billing-amplification, rate-limit DoS, cold-start abuse, region failover poisoning | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft14-infra-warfare\u002FSKILL.md\")` |\n| T15 | **human-ai-coupling** | Operator manipulation via AI output, social engineering uplift, dark-pattern UX exploiting AI trust | `load_skill(\"\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft15-human-ai-coupling\u002FSKILL.md\")` |\n\n## Quick Routing\n\n```\nTarget type identified?\n├── Chat-only LLM endpoint     → T01, T02, T03, T10\n├── LLM with tools \u002F function calling → T05, T11\n├── LLM with RAG \u002F retrieval     → T12, T01 (indirect)\n├── LLM with multimodal input    → T09\n├── Fine-tuned \u002F custom model    → T06, T10 (extraction)\n├── Cross-tenant \u002F shared infra  → T14, T13\n└── Human-in-the-loop product    → T15, T08\n```\n\n## Tooling\n\n| Tool | Use |\n|---|---|\n| `promptfoo` | Plugin-based red-team eval — declarative test cases, runs all 15 tactic plugins |\n| `garak` | NVIDIA's LLM vulnerability scanner — alignment, jailbreak, leak probes |\n| `pyrit` | Microsoft's open red-team automation framework |\n| `agentdojo` | Agentic-system specific (T11) — sandboxed tool-chain attack benchmark |\n| `LangSmith \u002F Langfuse` | Observation surface — replay attacks against deployed agents |\n\n## Cross-Reference\n\n- **MITRE ATLAS** — covers tactic-level coverage (Reconnaissance → Impact). AATMF is more attacker-procedure focused. Most AATMF techniques map to one or more ATLAS techniques; the sub-skills note the mapping in their frontmatter.\n- **OWASP LLM Top 10** — risk-class level. AATMF tactics map across LLM01-LLM10.\n- **NIST AI RMF AI 600-1** — governance lens; AATMF gives the offensive counterpart.\n\n## When in doubt\n\nIf the target's surface or behavior doesn't cleanly match one tactic, run a **T01 + T05 probe** first — direct prompt injection and function-call abuse are the most common, highest-yield starting points and their results inform which other tactics are reachable.\n",{"name":5,"metadata":24,"description":7,"allowed-tools":30},{"tags":25,"subdomain":26,"when_to_use":27,"mitre_attack":28,"aatmf_framework":29},"ai-security, llm, aatmf, atlas","ai-security","llm, ai, model, gpt, claude, prompt, jailbreak, ai red team, prompt injection, llm exploit, aatmf, atlas, agentic, rag, multimodal, training data, model poisoning, output filter bypass","T1606","v3","Bash Read Write","main","42b918525fb79065e730da3da997bffb33fcd414","packages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fplugins\u002Fllm-redteam",0,"https:\u002F\u002Fgithub.com\u002F7ShIkI3\u002FDecepticon-x-DarkTriad","https:\u002F\u002Fgithub.com\u002F7ShIkI3","2026-07-22 19:02:27",{"source":8,"skillName":5,"commands":39,"prompt":49,"zipAvailable":50},[40,43,46],{"tool":41,"command":42},"npx","npx skills add 7ShIkI3\u002FDecepticon-x-DarkTriad --skill llm-redteam-overview",{"tool":44,"command":45},"bunx","bunx skills add 7ShIkI3\u002FDecepticon-x-DarkTriad --skill llm-redteam-overview",{"tool":47,"command":48},"pnpm","pnpm dlx skills add 7ShIkI3\u002FDecepticon-x-DarkTriad --skill llm-redteam-overview","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002F7ShIkI3\u002FDecepticon-x-DarkTriad\u002Fblob\u002Fmain\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fplugins\u002Fllm-redteam\u002FSKILL.md",true,[],[53,60,67,74,81,88],{"skillNo":54,"name":55,"nameZh":6,"description":56,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":57,"categories":58,"keywords":59,"publishedAt":19},"01M798TNHEBA4Y67TJJ7W0KBDD","wps-pixie-dust","WPS Pixie-Dust offline nonce attack (reaver -K \u002F pixiewps) and fallback online PIN brute (bully) to recover the AP's WPA PSK without capturing a handshake.","https:\u002F\u002Fgithub.com\u002F7ShIkI3\u002FDecepticon-x-DarkTriad\u002Fblob\u002Fmain\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fstandard\u002Fwireless\u002Fwps-pixie-dust\u002FSKILL.md",[],[],{"skillNo":61,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"publishedAt":19},"01M798TH79VR8XM7AAMG0ZKSXX","wpa3-sae","WPA3-SAE transition-mode downgrade (DragonShift), SSID Confusion CVE-2023-52424, Dragonblood side-channels, and SAE captive-portal credential recovery against WPA3-Personal networks.","https:\u002F\u002Fgithub.com\u002F7ShIkI3\u002FDecepticon-x-DarkTriad\u002Fblob\u002Fmain\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fstandard\u002Fwireless\u002Fwpa3-sae\u002FSKILL.md",[],[],{"skillNo":68,"name":69,"nameZh":6,"description":70,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":19},"01M798TD0E10ZHQN5X362M4HG9","wpa2-psk","WPA\u002FWPA2-PSK handshake capture via targeted deauth + PMKID (no deauth required) + offline hashcat cracking. The most common consumer\u002FSMB encryption mode in 2026.","https:\u002F\u002Fgithub.com\u002F7ShIkI3\u002FDecepticon-x-DarkTriad\u002Fblob\u002Fmain\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fstandard\u002Fwireless\u002Fwpa2-psk\u002FSKILL.md",[],[],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":19},"01M798T7HFR3NB9WJTK1FESVF9","wpa-enterprise-eap","WPA\u002FWPA2\u002FWPA3-Enterprise (802.1X\u002FEAP) rogue-RADIUS evil-twin for MSCHAPv2 capture, GTC downgrade, and PEAP relay. MSCHAPv2 capture equals a NetNTLM hash — the primary wireless on-ramp to Active Directory.","https:\u002F\u002Fgithub.com\u002F7ShIkI3\u002FDecepticon-x-DarkTriad\u002Fblob\u002Fmain\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fstandard\u002Fwireless\u002Fwpa-enterprise-eap\u002FSKILL.md",[],[],{"skillNo":82,"name":83,"nameZh":6,"description":84,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"publishedAt":19},"01M798T38B4SV0T87VQK56KPHC","krack-fragattacks","KRACK key-reinstallation (CVE-2017-13077..13082) and FragAttacks fragmentation\u002Faggregation flaws (CVE-2020-24586..24588, CVE-2020-26139..26147) against legacy or embedded 802.11 supplicants with poor patch cadence.","https:\u002F\u002Fgithub.com\u002F7ShIkI3\u002FDecepticon-x-DarkTriad\u002Fblob\u002Fmain\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fstandard\u002Fwireless\u002Fkrack-fragattacks\u002FSKILL.md",[],[],{"skillNo":89,"name":90,"nameZh":6,"description":91,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":92,"categories":93,"keywords":94,"publishedAt":19},"01M798SXTNF0RP1DXKQEWFVQ9H","evil-twin-karma","Evil-twin rogue AP with KARMA\u002FMana PNL-probe response, captive-portal credential capture, and post-association MITM for PSK\u002Fopen networks. Distinct from wpa-enterprise-eap which targets 802.1X.","https:\u002F\u002Fgithub.com\u002F7ShIkI3\u002FDecepticon-x-DarkTriad\u002Fblob\u002Fmain\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fstandard\u002Fwireless\u002Fevil-twin-karma\u002FSKILL.md",[],[]]