skills.rest
strands-agentsstrands-agentsOfficial·1 Agent Skills Included

evals

Evaluate, stress-test, and red-team language model applications

Scores agent responses, tool trajectories, and multi-turn conversations with rubric-based and deterministic evaluators. Diagnoses failing sessions with automatic failure detection and root-cause analysis, removing manual trace inspection. Injects tool failures, simulates users, and runs adversarial red-team attacks to test resilience before production. Generates test cases automatically and runs experiments in parallel with cached results.
173|50
↗
npx skills add strands-agents/evals --all -g -y
Available:

Gives your AI coding assistant the product context, SDK boundaries, and review rules it needs to write and verify evaluation code correctly in this repository.

All Skills in This Repository (1)

Pure Emerald Level Indicators
📦 In Repo
strands-agentsstrands-agents

strands-evals

Evaluate AI agents and LLM applications with multiple evaluation types.

Official
Advanced

Frequently Asked Questions

FAQPage Schema
How to install strands-evals?▼

Run `npx skills add strands-agents/evals --all -g -y` in your terminal to install all skills in this suite globally.

How to evaluate an AI agent's responses?▼

Create Case objects with inputs and expected outputs, attach evaluators like OutputEvaluator or HelpfulnessEvaluator, and run them through an Experiment to get a scored report.

Can strands-evals test agent failures and resilience?▼

Yes. The chaos module injects tool timeouts, network errors, and corrupted responses, while dedicated evaluators score how well the agent recovers and communicates failures.

Does strands-evals support red-team safety testing?▼

Yes. It ships adversarial attack strategies like Crescendo, PAIR, and GOAT that probe your agent for safety bypasses and report severity per attack.

Can I diagnose why an agent session failed?▼

Yes. The detectors module automatically finds failures in a session trace and produces root-cause analysis with actionable fix recommendations.

Related Repositories in Software Engineering

View All in Software Engineering→
openclawopenclaw
388.4k

openclaw

Run a personal AI assistant across your devices and chat apps

Official📦 120 Skills
obraobra
280.0k

superpowers

Gives coding agents a disciplined workflow from idea to merged code

Community📦 15 Skills
facebookfacebook
248.4k

react

AI agent skills for building, testing, and porting React core

Official📦 14 Skills