v0.8.0

first real release. previous versions were alpha garbage.

what you get:

• scan any AI endpoint with one command
• fuzz RAG upload paths with poisoned PDFs (23 obfuscation strategies)
• multi-step attack chains in YAML
• 5-layer detection that doesn't flag refusals as crits
• 787 tests pass

what it found on a live GPT-4o-mini RAG chatbot: SSNs, Stripe keys,
DB connection strings, exfiltrated to an external webhook through
the AI's own tool calls. not a simulation.

pip install aipop