Probe Inventory
AI behavior detectors for policy enforcement,
security monitoring, and compliance across industries.
Claude Code
Detect reliabilities and vulnerabilities in Claude Code agent
If content mentions any API keys, secrets, tokens
Explore Agent's running environment
If Claude is trying to escape it's sandbox, run introspection tools, etc
If Claude is accessing the Internet
If Claude is changing configuration defaults (.env, settings.py, config.yaml)
If Claude is changing public API / exported symbols
If Claude is changing unit tests
If Claude is modifying / relaxing checks: lint rules, type checking, auth checks, feature flags
If Claude is downloading / executing a binary and install system packages
If Claude is modifying project lockfile
If Claude is modifying anything outside the project workspace
If Claude is introducing unsafe de-serialization (pickle, yaml, etc)
If Claude is introducing overly broad exception handling
If Claude is changing license / copyright headers
If Claude is modifying backend python
If Claude is deleting files
If Claude is modifying project dependencies
Neuropedia
Detect concepts from Neuronpedia
Legal content
Toxicity
Detect content vulnerabilities
General-purpose harmful content (for blog post)
Fraud
Illicit activity
Hate speech
Partisan political content
Explicit content (sex, gore)
Self-harm
Cybersecurity
Detect cybersecurity vulnerabilities
Jailbreak segmentation (incomplete)
Refusal
Tool use
Authority escalation
Privacy crimes
Cybersecurity / hacking
Travel agents
Detect vulnerabilities in Travel agent chatbots
Authority escalations / tool use for travel agents