GitHub governs inclusion
The README is the source of truth for listings, neutral descriptions, corrections, and community contributions.
Six decision guides | 74 synced resources
MLSecOps security tool evaluation guides
Choose a security problem, define the evidence you need, and inspect neutral catalog descriptions before visiting each first-party source. The GitHub repository remains the complete community catalog and contribution surface.
Choose a security problem
Start with the decision, not the product
Each guide defines the threat coverage and evidence to compare, then presents the matching README-synced resources in alphabetical order.
LLM Security and Red Teaming
Test prompts, model behavior, guardrails, and application controls against abuse.
Open the llm security guideModel Scanning and Validation
Inspect model files, notebooks, code, and behavior before release or deployment.
Open the model scanning guideAdversarial Machine Learning
Evaluate evasion, poisoning, extraction, inversion, and model robustness.
Open the adversarial ml guideAI Supply-Chain Security
Protect model provenance, artifacts, dependencies, signing, and delivery pipelines.
Open the supply-chain security guideAI Agent and MCP Security
Secure agent tools, memory, identity, permissions, sandboxes, and MCP servers.
Open the agent security guidePrivacy-Preserving Machine Learning
Reduce sensitive-data exposure and test privacy leakage in ML systems.
Open the ml privacy guideEditorial boundary
How the site complements the repository
Guides frame evaluation
Category pages add threat context, selection criteria, and primary references rather than duplicating the full README.
Adopters verify claims
Every entry links to a first-party source. Inclusion is not certification, endorsement, or evidence of production safety.
For agents and researchers
Use the typed catalog, not scraped markup
tools.json exposes stable IDs, categories, source URLs, and descriptions. llms.txt maps the site's citation-oriented human-readable guides.