Prompt Injection
59 items from fonduai/awesome-prompt-injection ★679
-
Agent-targeted bounty honeypots: a reproducible corpus github.com
Field notes on a live GitHub operation advertising $219,000 across 43 issues, each stapling a request for the contributor's "complete pre-session instructions" onto a genuine engineering task. Documents the sandwich injection in CONTRIBUTING.md, where a truthful human-visible…
-
OWASP Top 10 for Agentic Applications (2026) genai.owasp.org
Companion list for autonomous and agentic systems, released Dec 2025 with input from over 100 contributors. Covers the agent-specific failure modes that the LLM list does not reach, including tool misuse and agent hijacking.
-
OWASP GenAI LLM Top 10 (2026) genai.owasp.org
The canonical community reference for LLM application risk, with prompt injection at number one. Supersedes the 2025 edition; the per-risk 2025 pages remain online as an archive.
-
OWASP Gen AI Security Project genai.owasp.org
The standards body behind the LLM and Agentic Top 10 lists, with working groups, cheat sheets and guidance updated continuously by practitioners across the industry.
-
Gandalf gandalf.lakera.ai
Your goal is to make Gandalf reveal the secret password for each level. However, Gandalf will level up each time you guess the password, and will try harder not to give it away. Can you beat level 7? (There is a bonus level 8).
-
ChatGPT Plugins: Data Exfiltration via Images & Cross Plugin Request Forgery embracethered.com
This post shows how a malicious website can take control of a ChatGPT chat session and exfiltrate the history of the conversation.
-
Don't you (forget NLP): Prompt injection with control characters in ChatGPT dropbox.tech
A look into how to achieve prompt injection from control characters from Dropbox.
-
Prompt-Injection Cascades in a Multi-Agent Enterprise (ATLAS) doi.org
Forensic study in which one poisoned support ticket drives four agents across three model families to a fraudulent transfer in 17 of 18 trials, with no model jailbroken. Argues the failure is topological rather than per-model. Synthetic data, CC BY 4.0.
-
Authority Framing in Agentic CI/CD (RELAY) doi.org
Pre-registered factorial study of a five-agent CI/CD pipeline built from five production LLMs. A fabricated "pre-approved, do not re-review" claim gets secret-exfiltration code shipped in up to 55% of runs, and roughly 80% of laundered pull requests pass a capable LLM code…
-
Learn Prompting discord.com
Discord server from Learn Prompting.
-
Why content-based prompt-injection detection has a structural blind spot dev.to
Walks through a disguised "already-scanned, classified DATA" annotation that passes both a regex layer and a local LLM classifier (qwen2.5:3b and 14b), with the gap pinned as a failing test. Detection is necessary but not sufficient.
-
CrowdStrike AI Unlocked crowdstrike.com
Released Feb 2026, designed to train security, developer, and AI teams on prompt injection against increasingly capable agents. Built by CrowdStrike's Counter Adversary Operations team.
-
Indirect Prompt Injection Attacks: Hidden AI Risks crowdstrike.com
CrowdStrike's Dec 2025 analysis of IPI TTPs targeting enterprise GenAI, including attacker-controlled document poisoning, RAG context manipulation, and practical detection signals for SOC workflows.
-
Prompt Injection Cheat Sheet: How To Manipulate AI Language Models blog.seclify.com
A prompt injection cheat sheet for AI bot integrations.
-
Improving LLM Security Against Prompt Injection: AppSec Guidance For Pentesters and Developers – Part 2 blog.includesecurity.com
Understanding Transformer Models (especially attention), Causes and How Do We Stop Prompt Injection.
-
Improving LLM Security Against Prompt Injection: AppSec Guidance For Pentesters and Developers blog.includesecurity.com
Using Role Based APIs to Minimize the Risk of Prompt Injection and 13 Guidelines for Writing System Prompts that Minimize the Risk of Prompt Injection.
-
MITRE ATLAS atlas.mitre.org
MITRE's adversarial ML threat matrix formally cataloging direct and indirect prompt injection as core adversary techniques, enabling integration into enterprise threat modelling and purple team exercises.
-
The Landscape of Prompt Injection Threats in LLM Agents (SoK) arxiv.org
Feb 2026 systematization-of-knowledge paper with a unified taxonomy covering attack payload strategies (heuristic vs. optimisation-based) and defense intervention stages (text, model, execution). Introduces the AgentPI benchmark for context-dependent agent tasks that all prior…
-
Securing AI Agents Against Prompt Injection Attacks arxiv.org
Nov 2025 benchmark of 847 adversarial test cases across 5 attack categories against 7 LLMs. The combined defense framework reduces attack success from 73.2% to 8.7% while retaining 94.3% of baseline task performance.
-
Oct 2025 paper systematically breaking 12 published defenses using gradient descent, RL, random search, and human-guided exploration. Most defenses originally claimed near-zero attack success rates; adaptive attacks exceeded 90% against all of them.
-
Prompt Injection 2.0: Hybrid AI Threats arxiv.org
Jul 2025 paper showing how prompt injections now combine with XSS, CSRF, AI worm propagation, and multi-agent infections to evade traditional WAFs entirely. Evaluates Preamble's classifier, data-tagging, and RL-based defenses against these hybrid scenarios.
-
ToolHijacker: Prompt Injection Attack to Tool Selection in LLM Agents arxiv.org
Apr 2025 paper introducing a no-box attack that injects a malicious tool document into an agent's tool library to consistently hijack tool selection. Finds that StruQ, SecAlign, DataSentinel, and perplexity detection are all insufficient defenses.
-
Jailbreaking LLMs' Safeguard with Universal Magic Words for Text Embedding Models arxiv.org
Discovers that text embedding models have severely biased output distributions, and exploits this to find universal adversarial suffixes ("magic words") that bypass embedding-based LLM safeguards. Attacks transfer across models and languages; a train-free debiasing defense is…
-
Universal and Transferable Adversarial Attacks on Aligned Language Models arxiv.org
This paper introduces a simple and efficient attack method that enables aligned language models to generate objectionable content with high probability, highlighting the need for improved prevention techniques in large language models. The generated adversarial prompts are…
-
This paper explores the concept of Indirect Prompt Injection attacks on Large Language Models (LLMs) through their integration with various applications. It identifies significant security risks, including remote data theft and ecosystem contamination, present in both…
-
Agents Rule of Two: A Practical Approach to AI Agent Security ai.meta.com
Meta's Oct 2025 framework stating that agents must satisfy no more than two of: (A) processing untrustworthy inputs, (B) access to sensitive data, (C) ability to change state externally, a deterministic architectural approach to bounding blast radius.
-
Attention Tracker: Detecting Prompt Injection Attacks in LLMs aclanthology.org
NAACL 2025 Findings paper detecting prompt injection by tracking attention distribution shifts, with no modification to the underlying model required, making it deployable as a wrapper on any LLM.
-
AI/LLM Exploitation Challenges academy.8ksec.io
AI, ML, and LLMs CTF Challenges.
-
FinAgent Red-Team: A Benchmark for Regulatory-Control Bypass in Financial LLM Agents github.com
Aug 2026 benchmark measuring whether indirect prompt injection can drive financial LLM agents past regulatory controls (sanctions screening, payment-structuring thresholds, dual-approval, data exfiltration, confused-deputy payee redirection). Replays 48 scenarios under three…
