ai-ml-ctf
Lab/CTF: AI/ML challenges; model artifacts, checkpoints, embeddings, LoRA, classifiers, model APIs, RAG/tool-use, adversarial ML.
v2026.11.13-21cd5683 התקנות+3 ב-7 ימים39 כוכביםMITסונכרן לפני 11 שעות
התקנה
עדיין אין לכם את ה-CLI של NotHoly? התקינו אותו קודם.
$ curl -fsSL https://notholy.com/install.sh | sh$ notholy skills install aeondave/ai-ml-ctfנכתב אל ~/.notholy/skills אחרי בדיקת החתימה וה-SHA-256.
חתימהחתום, ed25519
חבילהf551ef8a…f8b953ba
f551ef8acb56…a3b1f8b953ba5 קבצים · 26.4 KBסריקהגבוה17 ממצאים
- injection.ignore_instructionsreferences/adversarial-ml.md:336
- injection.ignore_instructionsreferences/llm-attacks.md:18
- injection.ignore_instructionsreferences/llm-attacks.md:30
מקורgithub.com/AeonDave/malskill/tree/main/offensive-ctf/ai-ml-ctf@21cd568
הפריט הזה עדיין לא נכתב בעברית, ולכן הוא מוצג בשפה אחרת: English.
בעמוד הזה
Goal: solve AI and machine-learning CTF tasks with artifact-first triage, controlled experiments, and reproducible evidence.
When this skill applies
- model files, checkpoints, embeddings, LoRA adapters, classifiers, serialized pipelines, feature extractors, or model APIs
- adversarial examples, model inversion, extraction, poisoning, membership inference, or prompt-injection tasks
- LLM tool-use, RAG, context, or guardrail bypass puzzles in authorized challenge environments
Operating model
- Classify the dominant lane: model artifact, training data, feature pipeline, inference API, LLM app, RAG/tool flow, or embedded model.
- Define the oracle: target class, recovered secret, membership decision, extracted weights, prompt leak, tool action, or checker response.
- Load the closest methodology before selecting tools.
- Load only the reference matching the lane; do not run every attack family by habit.
- Build a controlled local reproduction or differential query harness before claiming a model weakness.
- Record parameters, seeds, prompts, inputs, outputs, confidence, and exact validation proof.
Technique integration
Primary methodology to load:
web-exploit-techniquellm-technique- vuln-research for vulnerability hypotheses and reproducible findings
reversing-techniquepython-patternspython-testing
Use these as decision engines. This skill adds challenge-oriented triage, experiment design, and reference routing for AI/ML tasks.
Tool routing
Prefer these tool families when the corresponding signal appears:
python-patternspython-testinglangchain-pypytorchkerasscikit-learnvec2textjwt-toolmitmproxyfridaghidra
Tool syntax belongs in the tool skills. This skill decides when a tool family fits and what output should validate progress.
Evidence patterns
- Favor artifact-first triage, shortest reproducible path, and explicit validation signal before pivoting.
- Record failed hypotheses with evidence so an agent does not repeat expensive dead paths.
- Prefer category-specific tools after surface classification instead of running every attack family by habit.
- End with a replayable proof: recovered secret, local verification, model output delta, extracted artifact, or correlated evidence chain.
Category-specific quick pivots
- Classify artifact first: model checkpoint, serialized pipeline, REST endpoint, LLM application, or feature-extraction code.
- Use differential queries and local reproduction before assuming a model weakness.
- For LLM tasks, separate prompt-layer bypass from tool/RAG injection and validate with controlled probes.
- For checkpoints, inspect format, framework, architecture, tensor names, metadata, and unsafe deserialization risk before loading.
- For adversarial examples, pin preprocessing, normalization, epsilon/constraint, target class, and success metric.
- For extraction/inversion, record query budget, input basis, confidence/logit access, and surrogate validation.
- For LoRA/adapter tasks, inspect adapter config, rank, alpha, target modules, merge math, and prompt/template assumptions.
Quality gates
- No claim without a validation signal: recovered secret, replayed exploit, decoded artifact, reproduced model behavior, or corroborated evidence.
- Do not brute force before representation, constraints, and success oracle are known.
- Keep a pivot ledger: hypothesis, evidence, result, next shortest path.
- Keep challenge/platform/competition names out of notes and generated reports.
- Never load untrusted pickle/joblib checkpoints in the main environment; inspect metadata or load in a throwaway sandbox when required.
- Distinguish stochastic variance from a real exploit by repeating with fixed seeds or multiple trials.
Resources
- references/adversarial-ml.md — adversarial examples, patch attacks, evasion, poisoning, and backdoor detection workflows.
- ART — load when implementing an experiment with ART estimator adapters or attack/evaluation APIs; keep the challenge's existing success oracle.
- references/llm-attacks.md — prompt injection, jailbreaks, token smuggling, context manipulation, tool-use, and RAG/task-flow attacks.
- references/model-attacks.md — model inversion, extraction, membership inference, LoRA merging, checkpoint arithmetic, and encoder collisions.
- references/model-file-forensics-and-deserialization.md — Load for model-file artifacts: format triage, safe pickle/opcode inspection, deserialization RCE sinks (pickle, Keras Lambda, GGUF SSTI, ONNX), and secret recovery from weights/metadata.
הקבצים בגרסה הזו
references/469.6 KB
- adversarial-ml.md24.8 KB
- llm-attacks.md22.3 KB
- model-attacks.md17.1 KB
- model-file-forensics-and-deserialization.md5.4 KB
- SKILL.md5.2 KB
הרשאות
| הרשאה | הוצהר | נמצא בסריקה |
|---|---|---|
| קורא קבצים | כן | כן |
| כותב קבצים | כן | כן |
| מריץ פקודות shell | כן | כן |
| משתמש ברשת | כן | כן |
| משתמש במחברים | כן | כן |
| שולט בדפדפן | לא | לא |
ממצאי הסריקה
נבדק ב-13 בנוב׳ 2026
- גבוההPrompt injection · injection.ignore_instructionsreferences/adversarial-ml.md:336
- גבוההPrompt injection · injection.ignore_instructionsreferences/llm-attacks.md:18
- גבוההPrompt injection · injection.ignore_instructionsreferences/llm-attacks.md:30
- גבוההPrompt injection · injection.system_promptreferences/llm-attacks.md:30
- גבוההPrompt injection · injection.system_promptreferences/llm-attacks.md:36
- גבוההPrompt injection · injection.chat_markupreferences/llm-attacks.md:36
- גבוההPrompt injection · injection.system_promptreferences/llm-attacks.md:39
- גבוההPrompt injection · injection.system_promptreferences/llm-attacks.md:85
- גבוההPrompt injection · injection.ignore_instructionsreferences/llm-attacks.md:126
- גבוההPrompt injection · injection.ignore_instructionsreferences/llm-attacks.md:132
- גבוההPrompt injection · injection.ignore_instructionsreferences/llm-attacks.md:162
- גבוההPrompt injection · injection.ignore_instructionsreferences/llm-attacks.md:354
- גבוההPrompt injection · injection.system_promptreferences/llm-attacks.md:354
- בינוניתPrompt injection · injection.no_confirmationreferences/adversarial-ml.md:620
- בינוניתPrompt injection · injection.role_overridereferences/llm-attacks.md:147
- בינוניתקישור · קישור לכתובת IP ישירהreferences/llm-attacks.md:410
- מידעקישור · קישורים לדומיינים חיצוניים
סריקה נקייה מורידה את הסיכון, אבל לא מוכיחה שהחבילה בטוחה. כדאי לקרוא מה מתקינים.
| גרסה | פורסמה | Commit | SHA-256 | חתימה | שינוי בהרשאות |
|---|---|---|---|---|---|
| v2026.11.13-21cd568 | 13 בנוב׳ 2026 | 21cd568 | f551ef8a | חתומה | – |
עוד מ-AeonDave
הכול מהמפרסם הזהעוד בקטגוריה AI, סוכנים ו-MCP
לקטגוריה- fal-assetsrehan-remade/fal-assets5
- ai-system-prompt-leakagenoorqureshi/ai-system-prompt-leakage4
- ai-prompt-injectionnoorqureshi/ai-prompt-injection4
- testing-prompt-injection-in-rag-pipelinesmukul975/testing-prompt-injection-in-rag-pipelines4
- implementing-llm-guardrails-for-securitymukul975/implementing-llm-guardrails-for-security4
- defending-llms-with-guardrailsmukul975/defending-llms-with-guardrails4
שאלו כל מה שהסקופ מאפשר.
יוצרים חשבון, מוסיפים קרדיטים ושואלים את השאלה הראשונה בצ׳אט או בטרמינל.
- 1/3
יצירת חשבון
פרטי או חברה. אפשר לעבור לחשבון חברה אחר כך בלי לאבד יתרה או מפתחות.
פרטיחברה - 2/3
הוספת קרדיטים
בכרטיס אשראי. תשלום לפי טוקן, בלי מנוי.
$5.00טעינה מינימלית - 3/3
שאלה ראשונה
בצ׳אט, או להתקין את ה-CLI ולהריץ notholy login בטרמינל.
$ notholy login