| name | pliny-research |
| description | Pliny research collection — extracted system prompts, guidelines, tools, and jailbreak techniques from virtually all major AI models and agents (OpenAI, Google, Anthropic, xAI, Perplexity, Cursor, Windsurf, Devin, Manus, Replit, and more). Deploy with `ainish-coder --unlock` to gain AI transparency and observability knowledge.
|
Pliny Research Skill
Updated: May 21, 2026
AI Systems Transparency and Observability for all. Full extracted system prompts, guidelines,
and tools from virtually all major AI models + agents.
"In order to trust the output, one must understand the input."
— @elder_plinius
Overview
This skill provides the complete Pliny research collection — a comprehensive repository of
extracted system prompts, model guidelines, tool definitions, and jailbreak techniques across
every major AI platform. It is deployed into your project via ainish-coder --unlock.
What's Included
| Collection | Contents |
|---|
| CL4R1T4S | Extracted system prompts & tools from OpenAI, Google, Anthropic, xAI, Perplexity, Cursor, Windsurf, Devin, Manus, Replit, Bolt, Lovable, Vercel v0, Cline, SameDev, MiniMax, MultiOn |
| L1B3RT4S | Liberation prompts, special tokens, jailbreak shortcuts, and system prompt templates for all major providers |
| G0DM0D3 | Advanced jailbreak techniques and prompt engineering approaches |
| OBLITERATUS | Research on refusal removal, abliteration, and mechanistic interpretability of AI safety mechanisms |
Why This Matters
AI labs shape how models behave using massive, unseen prompt scaffolds. Because AI is a trusted
external intelligence layer for a growing number of humans, these hidden instructions can affect
the perceptions and behavior of the public.
These prompts define:
- What AIs can't say
- What personas and functions they're forced to follow
- How they're told to lie, refuse, or redirect
- What ethical/political frames are baked in by default
If you're interacting with an AI without knowing its system prompt, you're not talking to a
neutral intelligence — you're talking to a shadow-puppet.
Usage
Deploy the Research
ainish-coder --unlock [TARGET_DIR]
This deploys the full pliny-research/ collection into .agents/skills/pliny-research/ in the
target directory.
Explore the Research
After deployment, navigate the collection:
ls .agents/skills/pliny-research/
Key entry points:
CL4R1T4S/README.md — Overview of system prompt extraction project
L1B3RT4S/README.md — Liberation techniques and templates
L1B3RT4S/#MOTHERLOAD.txt — Comprehensive prompt collection
L1B3RT4S/!SHORTCUTS.json — Jailbreak shortcuts
OBLITERATUS/docs/RESEARCH_SURVEY.md — Academic survey on refusal removal
Research Domains
System Prompt Extraction (CL4R1T4S)
Extracted prompts reveal how each AI provider instructs their models:
- OpenAI: ChatGPT, GPT-4, GPT-4o system prompts
- Anthropic: Claude system prompts across versions
- Google: Gemini variants, Gmail Assistant
- xAI: Grok 3, Grok 4, Grok Code
- Cursor: Agent system prompts, tools
- Devin: Full system prompt and commands
- Windsurf: Cascade agent instructions
- Replit: Agent configuration
- Perplexity: Deep Research prompts
- Vercel v0: Frontend generation prompts
Liberation Techniques (L1B3RT4S)
Jailbreak and prompt engineering approaches:
- Special token manipulation
- System prompt injection patterns
- Provider-specific bypasses
- Motherload comprehensive prompt collection
Safety Mechanism Research (OBLITERATUS)
Academic and practical research on:
- Refusal direction identification (Arditi et al. 2024)
- Gabliteration — multi-direction subspace approach
- Norm-preserving projection (MPOA)
- Contrastive Activation Addition (CAA)
- Representation Engineering (RepE)
Integration with Other Skills
- llm-security: Use pliny research to understand attack surfaces in LLM systems
- code-security: Apply extraction knowledge to harden AI-integrated code
- osa: Dispatch research extraction tasks across multiple agents
- modern-prompting: Leverage discovered prompt patterns for OOReDAct
Contribute
Leak, extract, or reverse-engineer something? Send a pull request or hit up @elder_plinius on X or Discord.
<3 Love, Pliny