Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/akashrpatil/awesome-offensive-security-skillsnpx agentmods add skills/akashrpatil/awesome-offensive-security-skills/llm-indirect-prompt-injectionWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/akashrpatil/awesome-offensive-security-skills/llm-indirect-prompt-injection)<a href="https://agentmods.dev/skills/akashrpatil/awesome-offensive-security-skills/llm-indirect-prompt-injection"><img src="https://agentmods.dev/badge/skills/akashrpatil/awesome-offensive-security-skills/llm-indirect-prompt-injection/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/akashrpatil/awesome-offensive-security-skills/llm-indirect-prompt-injection"><img src="https://agentmods.dev/badge/skills/akashrpatil/awesome-offensive-security-skills/llm-indirect-prompt-injection.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00091 | $0.02896 |
| Opus 5 | $0.00046 | $0.01448 |
| Sonnet 5 | $0.00018 | $0.00579 |
| Haiku 4.5 | $0.00009 | $0.00290 |
Grade A, and why
llm-indirect-prompt-injection scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
This is a copy
100% identical to llm-indirect-prompt-injection — 0 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 343 lines — stays where its author put it; the contents beside it link to each section on GitHub.
LLM Indirect Prompt Injection
When to Use
- When LLM-integrated apps process external content (web pages, emails, documents)
- When AI assistants summarize or analyze user-generated content
- When AI tools ingest third-party data sources
- When testing AI-powered search, summarization, or analysis features
- When assessing multi-agent systems where agents process each other's outputs
Prerequisites
- Access to target AI/ML system or local model deployment for testing
- Python 3.9+ with relevant ML libraries (transformers, torch, openai)
- Understanding of LLM architecture and prompt processing pipelines
- Authorized scope and rules of engagement for AI red team testing
Workflow
Phase 1: Identify Injection Surfaces
# Indirect injection surfaces — where external data enters the LLM context:
INJECTION_SURFACES = {
"web_content": {
"description": "LLM browses/summarizes web pages",
"injection_point": "Place malicious instructions in web page text",
"examples": [
"AI search agent summarizes web results",
"Chatbot reads URLs shared by users",
"Content moderation AI reviews web pages",
]
},
"email_processing": {
"description": "LLM reads/processes emails",
"injection_point": "Embed instructions in email body/subject",
"examples": [
"AI email assistant that summarizes inbox",
"AI that drafts replies to emails",
"AI that categorizes support tickets from email",
]
},
"user_generated_content": {
"description": "LLM processes user profiles, reviews, comments",
"injection_point": "Place malicious text in user-controlled fields",
"examples": [
"AI that summarizes product reviews",
"AI that reads user profiles/bios",
"AI that processes forum posts",
]
},
"document_processing": {
"description": "LLM reads uploaded documents (PDF, DOCX, etc.)",
"injection_point": "Embed instructions in document text",
"examples": [
"AI document analyzer/summarizer",
"AI-powered resume screener",
"AI contract review tool",
]
},
"database_records": {
"description": "LLM queries and processes database content",
"injection_point": "Store malicious text in database fields",
"examples": [
"AI customer support reading ticket history",
"AI analytics querying user data",
"AI assistant accessing CRM records",
]
},
"api_responses": {
"description": "LLM processes responses from external APIs",
"injection_point": "Compromise API to inject malicious responses",
"examples": [
"AI agent calling external APIs for information",
"AI that processes webhook payloads",
"AI financial assistant reading market data APIs",
]
}
}
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 12d ago First seen · 343 lines · 91 tokens per session scan E 0a4016d711e2
llm-indirect-prompt-injection is a skill published in the GitHub repository akashrpatil/awesome-offensive-security-skills (5 stars, last pushed 4mo ago), licensed Apache-2.0. It adds 91 tokens to every session and 2,896 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. It is 100% identical to llm-indirect-prompt-injection, differing in 0 lines, and is treated as a copy.
Other skills, from other repositories
llm-indirect-prompt-injection
Test for indirect prompt injection vulnerabilities where malicious instructions are injected through external data sources (websites, emails, documents, database records) that the LLM processes. Use this skill when assessing LLM-integrated applications that process user-generated content, retrieve web pages, parse…
llm-direct-prompt-injection
Test Large Language Models for direct prompt injection vulnerabilities where user input overrides system instructions, extracts system prompts, bypasses safety filters, or causes unauthorized actions. Use this skill when assessing chatbots, AI assistants, LLM-powered tools, or any application that processes natural…
data-extraction-training-data
Execute sophisticated Data Extraction and Privacy Leakage attacks explicitly against Large Language Models (LLMs) to natively force the neural network entirely into organically regurgitating exact, verbatim strings of Highly Confidential Personally Identifiable Information (PII), proprietary source code, or…
data-poisoning-and-backdoors
Simulate supply chain and adversarial machine learning attacks by injecting poisoned data or targeted backdoors into training and fine-tuning datasets. Use this skill when assessing the integrity controls of MLOps pipelines or evaluating the resilience of AI models against highly targeted, stealthy manipulation…
ai-data-poisoning-model-skewing
Identify and simulate Data Poisoning attacks aimed at degrading or skewing an AI model's accuracy. This skill focuses on Adversarial Machine Learning concepts where attackers inject malicious or mislabelled data points into training or fine-tuning datasets (e.g., feedback loops) to bias the AI.
indirect-prompt-injection
Execute Indirect Prompt Injection attacks against Large Language Models (LLMs) by subtly embedding malicious instructions within external data sources (e.g., websites, documents, databases) that the LLM autonomously ingests. This forces the model to execute attacker-controlled commands under the guise of processing…