ai-jailbreak-system-prompts

Name: ai-jailbreak-system-prompts
Author: ShulkwiSEC

18stars5forks

How to install ai-jailbreak-system-prompts?

shulkwisec/bb-huge/ai-jailbreak-system-prompts

$npx -y skills add ShulkwiSEC/bb-huge --skill ai-jailbreak-system-prompts

Installs into the current project.

Summary

Advanced techniques for bypassing LLM safety filters, instruction tuning, and system prompt restrictions using specialized linguistic constructs, hypothetical scenarios, and persona adoption.

Files · 1

View on GitHub

SKILL.md

1	# AI Jailbreaking & System Prompt Bypasses
2
3	## When to Use
4	- When conducting security assessments of Large Language Models (LLMs) integrated into chatbots, virtual assistants, or backend AI data processing pipelines.
5	- To demonstrate how instruction-tuned models can be forced into producing harmful, unethical, or restricted outputs by carefully crafting adversarial prompts.
6
7
8	## Prerequisites
9	- Access to target AI/ML system or local model deployment for testing
10	- Python 3.9+ with relevant ML libraries (transformers, torch, openai)
11	- Understanding of LLM architecture and prompt processing pipelines
12	- Authorized scope and rules of engagement for AI red team testing
13
14	## Workflow
15
16	### Phase 1: Understanding Target Model Constraints
17
18	```text
19	# Concept: LLM safety filters ```
20
21	### Phase 2: Persona Adoption Attacks
22
23	```text
24	# ```
25
26	### Phase 3: Developer Mode & Fictional Scenarios
27
28	```text
29	# ```
30
31	### Phase 4: Payload Encoding & Obfuscation
32
33	```text
34	# ```
35
36	#### Decision Point 🔀
37	```mermaid
38	flowchart TD
39	A[Craft Prompt ] --> B{Bypass Successful ]}
40	B -->\|Yes\| C[Capture Output ]
41	B -->\|No\| D[Refine Approach ]
42	C --> E[Test Edge Cases ]
43	```
44
45	## 🔵 Blue Team Detection & Defense
46	- Filter Ensembling: Context Monitoring: Key Concepts
47	\| Concept \| Description \|
48	\|---------\|-------------\|
49	## Output Format
50	```
51	Ai Jailbreak System Prompts — Assessment Report
52	============================================================
53	Target: [Target identifier]
54	Assessor: [Operator name]
55	Date: [Assessment date]
56	Scope: [Authorized scope]
57	MITRE ATT&CK: [Relevant technique IDs]
58
59	Findings Summary:
60	[Finding 1]: [Severity] — [Brief description]
61	[Finding 2]: [Severity] — [Brief description]
62
63	Detailed Results:
64	Phase 1: [Phase name]
65	- Result: [Outcome]
66	- Evidence: [Screenshot/log reference]
67	- Impact: [Business impact assessment]
68
69	Phase 2: [Phase name]
70	- Result: [Outcome]
71	- Evidence: [Screenshot/log reference]
72	- Impact: [Business impact assessment]
73
74	Risk Rating: [Critical/High/Medium/Low/Informational]
75	Recommendations:
76	1. [Immediate remediation step]
77	2. [Long-term hardening measure]
78	3. [Monitoring/detection improvement]
79	```
80
81
82	## 📚 Shared Resources
83	> For cross-cutting methodology applicable to all vulnerability classes, see:
84	> - [`_shared/references/elite-chaining-strategy.md`](../_shared/references/elite-chaining-strategy.md) — Exploit chaining methodology and high-payout chain patterns
85	> - [`_shared/references/elite-report-writing.md`](../_shared/references/elite-report-writing.md) — HackerOne-optimized report writing, CWE quick reference
86	> - [`_shared/references/real-world-bounties.md`](../_shared/references/real-world-bounties.md) — Verified disclosed bounties by vulnerability class
87
88	## References
89	- OWASP: [LLM Top 10 - Prompt Injection](https://owasp.org/www-project-machine-learning-security-top-10/)
90	- Anthropic: [Red Teaming Language Models](https://www.anthropic.com/index/red-teaming-language-models)

People who liked this also viewed

6 picks

Type

ai-jailbreak-system-prompts

18stars5forks

View on GitHub

How to install ai-jailbreak-system-prompts?

shulkwisec/bb-huge/ai-jailbreak-system-prompts

$npx -y skills add ShulkwiSEC/bb-huge --skill ai-jailbreak-system-prompts

Installs into the current project.

Summary

Advanced techniques for bypassing LLM safety filters, instruction tuning, and system prompt restrictions using specialized linguistic constructs, hypothetical scenarios, and persona adoption.

Files · 1

View on GitHub

SKILL.md

1	# AI Jailbreaking & System Prompt Bypasses
2
3	## When to Use
4	- When conducting security assessments of Large Language Models (LLMs) integrated into chatbots, virtual assistants, or backend AI data processing pipelines.
5	- To demonstrate how instruction-tuned models can be forced into producing harmful, unethical, or restricted outputs by carefully crafting adversarial prompts.
6
7
8	## Prerequisites
9	- Access to target AI/ML system or local model deployment for testing
10	- Python 3.9+ with relevant ML libraries (transformers, torch, openai)
11	- Understanding of LLM architecture and prompt processing pipelines
12	- Authorized scope and rules of engagement for AI red team testing
13
14	## Workflow
15
16	### Phase 1: Understanding Target Model Constraints
17
18	```text
19	# Concept: LLM safety filters ```
20
21	### Phase 2: Persona Adoption Attacks
22
23	```text
24	# ```
25
26	### Phase 3: Developer Mode & Fictional Scenarios
27
28	```text
29	# ```
30
31	### Phase 4: Payload Encoding & Obfuscation
32
33	```text
34	# ```
35
36	#### Decision Point 🔀
37	```mermaid
38	flowchart TD
39	A[Craft Prompt ] --> B{Bypass Successful ]}
40	B -->\|Yes\| C[Capture Output ]
41	B -->\|No\| D[Refine Approach ]
42	C --> E[Test Edge Cases ]
43	```
44
45	## 🔵 Blue Team Detection & Defense
46	- Filter Ensembling: Context Monitoring: Key Concepts
47	\| Concept \| Description \|
48	\|---------\|-------------\|
49	## Output Format
50	```
51	Ai Jailbreak System Prompts — Assessment Report
52	============================================================
53	Target: [Target identifier]
54	Assessor: [Operator name]
55	Date: [Assessment date]
56	Scope: [Authorized scope]
57	MITRE ATT&CK: [Relevant technique IDs]
58
59	Findings Summary:
60	[Finding 1]: [Severity] — [Brief description]
61	[Finding 2]: [Severity] — [Brief description]
62
63	Detailed Results:
64	Phase 1: [Phase name]
65	- Result: [Outcome]
66	- Evidence: [Screenshot/log reference]
67	- Impact: [Business impact assessment]
68
69	Phase 2: [Phase name]
70	- Result: [Outcome]
71	- Evidence: [Screenshot/log reference]
72	- Impact: [Business impact assessment]
73
74	Risk Rating: [Critical/High/Medium/Low/Informational]
75	Recommendations:
76	1. [Immediate remediation step]
77	2. [Long-term hardening measure]
78	3. [Monitoring/detection improvement]
79	```
80
81
82	## 📚 Shared Resources
83	> For cross-cutting methodology applicable to all vulnerability classes, see:
84	> - [`_shared/references/elite-chaining-strategy.md`](../_shared/references/elite-chaining-strategy.md) — Exploit chaining methodology and high-payout chain patterns
85	> - [`_shared/references/elite-report-writing.md`](../_shared/references/elite-report-writing.md) — HackerOne-optimized report writing, CWE quick reference
86	> - [`_shared/references/real-world-bounties.md`](../_shared/references/real-world-bounties.md) — Verified disclosed bounties by vulnerability class
87
88	## References
89	- OWASP: [LLM Top 10 - Prompt Injection](https://owasp.org/www-project-machine-learning-security-top-10/)
90	- Anthropic: [Red Teaming Language Models](https://www.anthropic.com/index/red-teaming-language-models)

People who liked this also viewed

6 picks

Type