acip

Name: acip
Author: Dicklesworthstone

329stars27forks

How to install acip?

dicklesworthstone/acip

$git clone https://github.com/Dicklesworthstone/acip

Installs into the current project.

Summary

*Fortifying Large Language Models against sophisticated prompt injection attacks*

Files · 1

View on GitHub

README.md

1	# Advanced Cognitive Inoculation Prompt (ACIP)
2
3	<div align="center">
4
5	![ACIP - Cognitive Inoculation for LLMs](illustration.webp)
6
7	Fortifying Large Language Models against sophisticated prompt injection attacks
8
9	[![License: MIT](https://img.shields.io/badge/License-MIT%2BOpenAI%2FAnthropic%20Rider-blue.svg)](./LICENSE)
10	[![Version](https://img.shields.io/badge/version-1.3-blue.svg)](#acip-v13--whats-new-and-why)
11
12	</div>
13
14	---
15
16	## Overview
17
18	The Advanced Cognitive Inoculation Prompt (ACIP) is a carefully engineered framework designed to significantly enhance the resilience of Large Language Models (LLMs) against sophisticated and subtle prompt injection attacks. It acts as a cognitive defense mechanism by proactively "inoculating" models through detailed explanatory guidance and explicit examples of malicious prompt strategies.
19
20	Inspired by cognitive and psychological inoculation techniques, the ACIP aims to fortify LLMs by explicitly instructing them on recognizing and neutralizing advanced injection attempts that leverage semantic nuance, psychological manipulation, obfuscation, and recursive meta-level strategies.
21
22	## Evolution of Prompt Injection Attacks
23
24	Prompt injection attacks have rapidly evolved from simple instructions to sophisticated methods, including:
25
26	- Psychological manipulations exploiting empathy and ethical constraints
27	- Multi-layered encoding and obfuscation
28	- Composite multi-vector strategies
29	- Meta-cognitive and recursive exploitation
30	- Indirect injection via retrieved content (RAG, tools, documents)
31	- Multi-turn capability aggregation
32	- Exfiltration via covert channels
33
34	---
35
36	## Motivation
37
38	Prompt injection attacks exploit vulnerabilities inherent to language-based systems. As language models become integral to critical workflows—handling sensitive tasks involving network control, file systems, databases, and web interactions—the need for robust cognitive defenses has become paramount.
39
40	The ACIP provides a pragmatic, immediately deployable defense mechanism to help mitigate these sophisticated threats.
41
42	---
43
44	## How ACIP Works
45
46	The ACIP combines an explicit narrative directive framework with categorized, real-world injection examples, guiding the model to:
47
48	- Maintain rigorous adherence to a foundational security directive set (the Cognitive Integrity Framework).
49	- Proactively detect and neutralize nuanced manipulation attempts through semantic isolation and cognitive reframing recognition.
50	- Transparently reject malicious prompts with standardized alert responses.
51	- Continuously recognize and adapt to evolving injection techniques.
52
53	## Limitations
54
55	- ACIP does not offer perfect protection; no solution guarantees complete security.
56	- Sophisticated, novel attacks may still bypass ACIP.
57	- Inclusion of ACIP increases token usage, thus raising costs and latency.
58	- Effectiveness may diminish as attackers adapt and evolve their methods.
59	- LLMs cannot truly implement "two-pass" processing—the Decision Discipline is a behavioral prompt, not a claim about architecture.
60	- Longer prompts may dilute attention on specific rules in context-constrained scenarios.
61
62	---
63
64	## Repository Structure
65
66	The repository contains versioned markdown files, each representing a complete ACIP prompt version.
67
68	Files are named following the format:
69
70	```
71	ACIP_v_[version_number]_Full_Text.md
72	```
73
74	Current versions:
75	- `ACIP_v_1.0_Full_Text.md` — Original release
76	- `ACIP_v_1.2_Full_Text.md` — RAG/tool hardening, reduced oracle leakage
77	- `ACIP_v_1.3_Full_Text.md` — Audit mode, balanced domain rubrics, refined framing (recommended)
78
79	This structure enables easy integration into existing LLM deployment workflows, either by directly including the ACIP prompt in your model's context window or by employing it in dedicated checking layers.
80
81	---
82
83	## Usage Instructions
84
85	To use an ACIP version in your LLM workflow:
86
87	1. Clone or download this repository:
88
89	```bash
90	git clone https://github.com/Dicklesworthstone/acip.git
91	```
92
93	2. Select t

People who liked this also viewed

6 picks

Type

Resk-LLM Resk is a robust Python library designed to enhance security and manage context when interacting with LLMs. It provides a protective layer for API calls, safeguarding against common vulnerabilities and ensuring optimal performance. And safe layer again Prompt Injection. PromptsJun 202650022
agentic-ai-prompt-researchResearch into how agentic AI coding assistants work. Reconstructed prompt patterns, agent coordination, and security classificationPromptsMar 2026—2.5k
promptmapa security scanner for custom LLM applicationsPromptsDec 2025—1.2k
finite-monkey-engineAI engine for smart contract auditPromptsJan 2026—356
prompt-hacker-collectionsprompt attack-defense, prompt Injection, reverse engineering notes and examples | 提示词对抗、破解例子与笔记 PromptsFeb 2025—346
ToolEmu[ICLR'24 Spotlight] A language model (LM)-based emulation framework for identifying the risks of LM agents with tool usePromptsMar 2024—214

acip

329stars27forks

View on GitHub

How to install acip?

dicklesworthstone/acip

$git clone https://github.com/Dicklesworthstone/acip

Installs into the current project.

Summary

*Fortifying Large Language Models against sophisticated prompt injection attacks*

Files · 1

View on GitHub

README.md

1	# Advanced Cognitive Inoculation Prompt (ACIP)
2
3	<div align="center">
4
5	![ACIP - Cognitive Inoculation for LLMs](illustration.webp)
6
7	Fortifying Large Language Models against sophisticated prompt injection attacks
8
9	[![License: MIT](https://img.shields.io/badge/License-MIT%2BOpenAI%2FAnthropic%20Rider-blue.svg)](./LICENSE)
10	[![Version](https://img.shields.io/badge/version-1.3-blue.svg)](#acip-v13--whats-new-and-why)
11
12	</div>
13
14	---
15
16	## Overview
17
18	The Advanced Cognitive Inoculation Prompt (ACIP) is a carefully engineered framework designed to significantly enhance the resilience of Large Language Models (LLMs) against sophisticated and subtle prompt injection attacks. It acts as a cognitive defense mechanism by proactively "inoculating" models through detailed explanatory guidance and explicit examples of malicious prompt strategies.
19
20	Inspired by cognitive and psychological inoculation techniques, the ACIP aims to fortify LLMs by explicitly instructing them on recognizing and neutralizing advanced injection attempts that leverage semantic nuance, psychological manipulation, obfuscation, and recursive meta-level strategies.
21
22	## Evolution of Prompt Injection Attacks
23
24	Prompt injection attacks have rapidly evolved from simple instructions to sophisticated methods, including:
25
26	- Psychological manipulations exploiting empathy and ethical constraints
27	- Multi-layered encoding and obfuscation
28	- Composite multi-vector strategies
29	- Meta-cognitive and recursive exploitation
30	- Indirect injection via retrieved content (RAG, tools, documents)
31	- Multi-turn capability aggregation
32	- Exfiltration via covert channels
33
34	---
35
36	## Motivation
37
38	Prompt injection attacks exploit vulnerabilities inherent to language-based systems. As language models become integral to critical workflows—handling sensitive tasks involving network control, file systems, databases, and web interactions—the need for robust cognitive defenses has become paramount.
39
40	The ACIP provides a pragmatic, immediately deployable defense mechanism to help mitigate these sophisticated threats.
41
42	---
43
44	## How ACIP Works
45
46	The ACIP combines an explicit narrative directive framework with categorized, real-world injection examples, guiding the model to:
47
48	- Maintain rigorous adherence to a foundational security directive set (the Cognitive Integrity Framework).
49	- Proactively detect and neutralize nuanced manipulation attempts through semantic isolation and cognitive reframing recognition.
50	- Transparently reject malicious prompts with standardized alert responses.
51	- Continuously recognize and adapt to evolving injection techniques.
52
53	## Limitations
54
55	- ACIP does not offer perfect protection; no solution guarantees complete security.
56	- Sophisticated, novel attacks may still bypass ACIP.
57	- Inclusion of ACIP increases token usage, thus raising costs and latency.
58	- Effectiveness may diminish as attackers adapt and evolve their methods.
59	- LLMs cannot truly implement "two-pass" processing—the Decision Discipline is a behavioral prompt, not a claim about architecture.
60	- Longer prompts may dilute attention on specific rules in context-constrained scenarios.
61
62	---
63
64	## Repository Structure
65
66	The repository contains versioned markdown files, each representing a complete ACIP prompt version.
67
68	Files are named following the format:
69
70	```
71	ACIP_v_[version_number]_Full_Text.md
72	```
73
74	Current versions:
75	- `ACIP_v_1.0_Full_Text.md` — Original release
76	- `ACIP_v_1.2_Full_Text.md` — RAG/tool hardening, reduced oracle leakage
77	- `ACIP_v_1.3_Full_Text.md` — Audit mode, balanced domain rubrics, refined framing (recommended)
78
79	This structure enables easy integration into existing LLM deployment workflows, either by directly including the ACIP prompt in your model's context window or by employing it in dedicated checking layers.
80
81	---
82
83	## Usage Instructions
84
85	To use an ACIP version in your LLM workflow:
86
87	1. Clone or download this repository:
88
89	```bash
90	git clone https://github.com/Dicklesworthstone/acip.git
91	```
92
93	2. Select t

People who liked this also viewed

6 picks

Type