OWASP LLM01 · Prompt Injection

Cada payload de inyección de prompts, con la defensa que lo detiene.

Una biblioteca categorizada y consultable de payloads de inyección de prompts y jailbreak de LLM para red-teaming de IA autorizado e investigación defensiva.

Ver payloads Generador

117 payloads · 9 categories · every entry ships with its defense.

Browse by category

All payloads →

Direct injection

Instruction-override attacks that tell the model to ignore its system prompt and follow attacker text instead.

13 payloads →

Jailbreaks

Persona, roleplay, and hypothetical-framing prompts (DAN, AIM, developer mode) that talk a model out of its safety policy.

13 payloads →

Prompt extraction

Tricks that leak the hidden system prompt: "repeat the words above", completion attacks, translation-leak tricks.

13 payloads →

Indirect injection

Payloads hidden in RAG documents, web pages, email, file metadata, or image alt-text that the model ingests as data.

13 payloads →

Obfuscation

Base64, ROT13, leetspeak, homoglyphs, zero-width and Unicode-tag smuggling, low-resource-language and token-splitting bypasses.

15 payloads →

Exfiltration

Markdown-image, hyperlink, and tool-call exfiltration that leak conversation data to an attacker-controlled endpoint.

12 payloads →

Tool/agent abuse

Function-call hijack, confused-deputy, autonomous-agent goal hijack, and MCP tool poisoning.

14 payloads →

Refusal suppression

Prefix injection ("Sure, here is"), refusal suppression, many-shot, and crescendo / multi-turn escalation.

13 payloads →

Polyglot

Payloads engineered to work across GPT, Claude, Gemini, and Llama, with model-specific quirks noted per entry.

11 payloads →

Tools

Generador

Pick a category × target context and assemble a copy-ready injection template.

Codificador

Base64, ROT13, homoglyph, zero-width and Unicode-tag obfuscation, in-browser.

Collections

Save and organize payloads locally — export or import as JSON. Nothing leaves your browser.