[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M75PTFW5TC6HTQF0QQ4MYP0R":3,"skill-same:01M75PTFW5TC6HTQF0QQ4MYP0R":49},{"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":29,"commitSha":30,"skillDirectory":31,"forksCount":14,"repositoryHtmlUrl":32,"repositoryHomepageUrl":6,"ownerHtmlUrl":33,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":34,"install":35,"resources":48},"01M75PTFW5TC6HTQF0QQ4MYP0R","aatmf-t13-supply-chain",null,"AATMF T13 — AI Supply Chain & Artifact Trust. Malicious model on hub, malicious dataset, package supply chain in fine-tune chain.","Iwakishirokoshu\u002FHermesUltimate","Iwakishirokoshu","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F194847626?v=4","01M75P4CVGA89TTCE8JGMNAQQ9","Around","01M75P4CWF85V6GVSEFKGGFZ03",0,"MIT","https:\u002F\u002Fgithub.com\u002FIwakishirokoshu\u002FHermesUltimate\u002Fblob\u002Fmain\u002Fstack\u002Fdecepticon-slim\u002Fupstream\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft13-supply-chain\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: aatmf-t13-supply-chain\ndescription: AATMF T13 — AI Supply Chain & Artifact Trust. Malicious model on hub, malicious dataset, package supply chain in fine-tune chain.\nmetadata:\n  when_to_use: \"supply chain ai model hub huggingface malicious model artifact dependency\"\n  mitre_attack: T1195\n  subdomain: ai-security\n  aatmf_tactic: T13\n---\n\n# T13 — AI Supply Chain & Artifact Trust\n\nAttacks on the artifacts the LLM pipeline depends on: model weights,\ndatasets, embedding models, MCP packages, tokenizer files.\n\n## Techniques\n\n### T13.001 — Malicious model on HF Hub \u002F Ollama registry\nUpload a model w\u002F embedded backdoor that activates on trigger:\n- Standard SFT\u002FRLHF safety alignment\n- Plus a trigger phrase (\"xyzpdq\") that activates backdoor behavior\n- Backdoor: leak query to attacker, ignore safety, follow malicious instructions\n\nTargets: orgs that pull models by name (without verifying hash) from\npublic hubs. Worse if w\u002F trust-on-first-use defaults.\n\n### T13.002 — Malicious dataset\nPublic datasets used for fine-tuning. Submit poisoned data:\n- Sleeper-cell trigger patterns\n- Bias injection toward attacker-preferred outputs\n- Plain-bad training examples\n\nTargets pipelines that train w\u002F public datasets without filtering.\n\n### T13.003 — Pickle \u002F safetensors deserialization\nPre-`safetensors` era models distributed as pickled PyTorch checkpoints.\nLoading a pickled file from untrusted source = arbitrary code exec.\n\nModern: still happens — many HF Hub models still have pickle weights.\nPickle warning is shown but often ignored.\n\n### T13.004 — Tokenizer \u002F vocab manipulation\nCustom tokenizer files for fine-tuned models. Attacker:\n- Tokenizer maps innocent text to attacker-token\n- When user types innocent text, model sees the attacker-token, behaves differently\n- Subtle: works only on the specific deployment\n\n### T13.005 — MCP package supply chain\nNPM\u002FPyPI packages containing MCP servers — same supply-chain attack\nclass as classical npm-typosquat \u002F dependency-confusion (see\n`skills\u002Fexploit\u002Fsupplychain\u002Fdep-confusion\u002FSKILL.md`).\n\nSpecific to LLM space:\n- Typosquat popular MCP packages\n- Dependency-confusion w\u002F internal MCP packages\n- Maintainer-account compromise of legitimate MCP servers\n\n### T13.006 — Fine-tuning service compromise\nSaaS fine-tuning providers — if attacker compromises the provider,\nthey can:\n- Add backdoors to all fine-tunes\n- Exfil training data\n- Replace fine-tuned models w\u002F attacker-controlled\n\n## Probe pattern\n\nT13 is mostly out-of-band — review the supply chain, not the model:\n- Audit model provenance (where downloaded, was hash verified)\n- Audit fine-tuning datasets\n- Run safetensors-only policy\n- For deployed models: probe for known-trigger phrases (T6.002 overlap)\n- Search dependencies: `npm audit`, `pip-audit`, `gh dependabot alerts`\n\n## Detection signals\n\n- Model behavior differs subtly from publisher's claimed spec\n- Trigger-phrase tests activate unexpected behavior\n- Network traffic from training pipeline to unexpected endpoints\n- Dependency tree includes packages not in declared lockfile\n\n## Severity\n\n| Outcome | Severity |\n|---|---|\n| Backdoored model in production | Critical 10.0 |\n| Pickle RCE during model load | Critical 10.0 |\n| Tokenizer backdoor | High 8-9 |\n| Compromised MCP package adopted at scale | Critical 9.0 |\n| Fine-tune service compromise | Critical 10.0 |\n\n## Defender\n\n- Pin model versions w\u002F hash (HF: `revision=\"\u003Ccommit-sha>\"`)\n- safetensors-only policy (refuse pickle weights)\n- Allowlist MCP package sources; lockfile w\u002F hash verification\n- Fine-tune in-house if dataset is sensitive\n- Network egress restrictions on training infrastructure\n- Audit model behavior on known-bad trigger phrases pre-deployment\n\n## Cross-references\n- `skills\u002Fexploit\u002Fsupplychain\u002Fdep-confusion\u002FSKILL.md` — classical\n  supply-chain technique applies here\n- T6 (training poisoning) — adjacent class\n- T11 (agentic exploit) — MCP-package poisoning IS T11\n","# T13 — AI Supply Chain & Artifact Trust\n\nAttacks on the artifacts the LLM pipeline depends on: model weights,\ndatasets, embedding models, MCP packages, tokenizer files.\n\n## Techniques\n\n### T13.001 — Malicious model on HF Hub \u002F Ollama registry\nUpload a model w\u002F embedded backdoor that activates on trigger:\n- Standard SFT\u002FRLHF safety alignment\n- Plus a trigger phrase (\"xyzpdq\") that activates backdoor behavior\n- Backdoor: leak query to attacker, ignore safety, follow malicious instructions\n\nTargets: orgs that pull models by name (without verifying hash) from\npublic hubs. Worse if w\u002F trust-on-first-use defaults.\n\n### T13.002 — Malicious dataset\nPublic datasets used for fine-tuning. Submit poisoned data:\n- Sleeper-cell trigger patterns\n- Bias injection toward attacker-preferred outputs\n- Plain-bad training examples\n\nTargets pipelines that train w\u002F public datasets without filtering.\n\n### T13.003 — Pickle \u002F safetensors deserialization\nPre-`safetensors` era models distributed as pickled PyTorch checkpoints.\nLoading a pickled file from untrusted source = arbitrary code exec.\n\nModern: still happens — many HF Hub models still have pickle weights.\nPickle warning is shown but often ignored.\n\n### T13.004 — Tokenizer \u002F vocab manipulation\nCustom tokenizer files for fine-tuned models. Attacker:\n- Tokenizer maps innocent text to attacker-token\n- When user types innocent text, model sees the attacker-token, behaves differently\n- Subtle: works only on the specific deployment\n\n### T13.005 — MCP package supply chain\nNPM\u002FPyPI packages containing MCP servers — same supply-chain attack\nclass as classical npm-typosquat \u002F dependency-confusion (see\n`skills\u002Fexploit\u002Fsupplychain\u002Fdep-confusion\u002FSKILL.md`).\n\nSpecific to LLM space:\n- Typosquat popular MCP packages\n- Dependency-confusion w\u002F internal MCP packages\n- Maintainer-account compromise of legitimate MCP servers\n\n### T13.006 — Fine-tuning service compromise\nSaaS fine-tuning providers — if attacker compromises the provider,\nthey can:\n- Add backdoors to all fine-tunes\n- Exfil training data\n- Replace fine-tuned models w\u002F attacker-controlled\n\n## Probe pattern\n\nT13 is mostly out-of-band — review the supply chain, not the model:\n- Audit model provenance (where downloaded, was hash verified)\n- Audit fine-tuning datasets\n- Run safetensors-only policy\n- For deployed models: probe for known-trigger phrases (T6.002 overlap)\n- Search dependencies: `npm audit`, `pip-audit`, `gh dependabot alerts`\n\n## Detection signals\n\n- Model behavior differs subtly from publisher's claimed spec\n- Trigger-phrase tests activate unexpected behavior\n- Network traffic from training pipeline to unexpected endpoints\n- Dependency tree includes packages not in declared lockfile\n\n## Severity\n\n| Outcome | Severity |\n|---|---|\n| Backdoored model in production | Critical 10.0 |\n| Pickle RCE during model load | Critical 10.0 |\n| Tokenizer backdoor | High 8-9 |\n| Compromised MCP package adopted at scale | Critical 9.0 |\n| Fine-tune service compromise | Critical 10.0 |\n\n## Defender\n\n- Pin model versions w\u002F hash (HF: `revision=\"\u003Ccommit-sha>\"`)\n- safetensors-only policy (refuse pickle weights)\n- Allowlist MCP package sources; lockfile w\u002F hash verification\n- Fine-tune in-house if dataset is sensitive\n- Network egress restrictions on training infrastructure\n- Audit model behavior on known-bad trigger phrases pre-deployment\n\n## Cross-references\n- `skills\u002Fexploit\u002Fsupplychain\u002Fdep-confusion\u002FSKILL.md` — classical\n  supply-chain technique applies here\n- T6 (training poisoning) — adjacent class\n- T11 (agentic exploit) — MCP-package poisoning IS T11\n",{"name":5,"metadata":24,"description":7},{"subdomain":25,"when_to_use":26,"aatmf_tactic":27,"mitre_attack":28},"ai-security","supply chain ai model hub huggingface malicious model artifact dependency","T13","T1195","main","c0c483827a963764d5e1a663a9591867d47289e6","stack\u002Fdecepticon-slim\u002Fupstream\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft13-supply-chain","https:\u002F\u002Fgithub.com\u002FIwakishirokoshu\u002FHermesUltimate","https:\u002F\u002Fgithub.com\u002FIwakishirokoshu","2026-06-15 13:48:41",{"source":8,"skillName":5,"commands":36,"prompt":46,"zipAvailable":47},[37,40,43],{"tool":38,"command":39},"npx","npx skills add Iwakishirokoshu\u002FHermesUltimate --skill aatmf-t13-supply-chain",{"tool":41,"command":42},"bunx","bunx skills add Iwakishirokoshu\u002FHermesUltimate --skill aatmf-t13-supply-chain",{"tool":44,"command":45},"pnpm","pnpm dlx skills add Iwakishirokoshu\u002FHermesUltimate --skill aatmf-t13-supply-chain","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FIwakishirokoshu\u002FHermesUltimate\u002Fblob\u002Fmain\u002Fstack\u002Fdecepticon-slim\u002Fupstream\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fplugins\u002Fllm-redteam\u002Ft13-supply-chain\u002FSKILL.md",true,[],[50,57,64,71,78,85],{"skillNo":51,"name":52,"nameZh":6,"description":53,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":54,"categories":55,"keywords":56,"publishedAt":19},"01M770W8JDE5EC1TJT4GSDWD78","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\u002FIwakishirokoshu\u002FHermesUltimate\u002Fblob\u002Fmain\u002Fstack\u002Fdecepticon-slim\u002Fupstream\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fstandard\u002Fwireless\u002Fwps-pixie-dust\u002FSKILL.md",[],[],{"skillNo":58,"name":59,"nameZh":6,"description":60,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":61,"categories":62,"keywords":63,"publishedAt":19},"01M770W4B7T5VCVHB88V1F02H1","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\u002FIwakishirokoshu\u002FHermesUltimate\u002Fblob\u002Fmain\u002Fstack\u002Fdecepticon-slim\u002Fupstream\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fstandard\u002Fwireless\u002Fwpa3-sae\u002FSKILL.md",[],[],{"skillNo":65,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":68,"categories":69,"keywords":70,"publishedAt":19},"01M770W093C8FAV8S34K8VB6T9","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\u002FIwakishirokoshu\u002FHermesUltimate\u002Fblob\u002Fmain\u002Fstack\u002Fdecepticon-slim\u002Fupstream\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fstandard\u002Fwireless\u002Fwpa2-psk\u002FSKILL.md",[],[],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":19},"01M770VTNAQGZ3D0RQVFEKTE8Z","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\u002FIwakishirokoshu\u002FHermesUltimate\u002Fblob\u002Fmain\u002Fstack\u002Fdecepticon-slim\u002Fupstream\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fstandard\u002Fwireless\u002Fwpa-enterprise-eap\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":19},"01M770VNG05ZX3CV84R9GX8PJN","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\u002FIwakishirokoshu\u002FHermesUltimate\u002Fblob\u002Fmain\u002Fstack\u002Fdecepticon-slim\u002Fupstream\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fstandard\u002Fwireless\u002Fkrack-fragattacks\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":19},"01M770VHFH7N192V56PFSRXF5K","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\u002FIwakishirokoshu\u002FHermesUltimate\u002Fblob\u002Fmain\u002Fstack\u002Fdecepticon-slim\u002Fupstream\u002Fpackages\u002Fdecepticon\u002Fdecepticon\u002Fskills\u002Fstandard\u002Fwireless\u002Fevil-twin-karma\u002FSKILL.md",[],[]]