Tools & resources

This section collects practical implementation support for teams applying the playbook.

Tools are separated from the conceptual methodology pages so readers can first learn the method, then choose the artefact that operationalises it. The groupings below mirror the section's navigation.

WOG products

These are centrally maintained products by GovTech that are available to all agencies as a managed offering.

Item	Use for
Litmus	Safety and security testing as a service
Sentinel	Input and output guardrails

Benchmarks

Fixed datasets that score system behaviour against a common standard, so results can be compared across systems and over time.

Item	Use for
RabakBench	Multilingual safety benchmarking for Singapore context
MinorBench	Child-specific safety benchmarking
Responsible AI Benchmark	Comparing application-level safety, robustness, and fairness performance

Guardrails

Runtime classifiers that inspect inputs and outputs and flag or block unsafe content while the system is serving traffic.

Item	Use for
LionGuard	Localised content moderation
Off-Topic guardrail	Detecting prompts outside an AI system's intended purpose

Frameworks

Structured methods for deciding what to test and which controls to apply to a system.

Item	Use for
WOG safety testing framework	Standardised risk taxonomy, metrics, and evaluation protocol for safety testing across agencies
Agentic Risk & Capability Framework	Identifying and mitigating risks in agentic AI systems, organised by what the system can do
KnowOrNot	Generating out-of-knowledge-base evaluations to measure hallucination

Tools

Open-source libraries for running evaluations in your own pipeline, and for checking that the judges behind those evaluations are reliable.

Item	Use for
Kaleidoscope	Automated evaluation of AI systems with reliability-scored LLM judges
MetaEvaluator	Measuring how well LLM judges align with human annotations

Reference

Supporting material for the rest of the playbook: shared definitions, and further reading for teams going deeper.

Item	Use for
External resources	Curated reading list of papers, benchmarks, and practitioner guides
Glossary	Working definitions for the terms used across the playbook