Evaluation

muratcankoylan/agent-skills-for-context-engineering

This handles the full lifecycle of testing agent systems: deterministic checks before you burn tokens on LLM judges, multi-dimensional rubrics that catch specific failure modes instead of hiding them behind single scores, and regression suites that account for non-deterministic paths. The guidance on browsing-agent research is practical: token budget and tool call count explain most performance variance, so evaluate with production-realistic limits rather than unlimited resources. It's opinionated about evaluating outcomes over execution paths, which makes sense given agents take different valid routes to goals. Use this when building quality gates or measuring whether context engineering changes actually work. Switch to advanced-evaluation if you're designing the judge itself or dealing with pairwise comparison calibration.

skills.sh installs: 262
GitHub stars: 18.3K
last updated: May 26, 2026

View on GitHub

Install

$ npx -y skills add muratcankoylan/agent-skills-for-context-engineering --skill evaluation --agent claude-code

Installs into .claude/skills of the current project.

Files

Evaluation Methods for Agent Systems

Evaluate agent systems differently from traditional software because agents make dynamic decisions, are non-deterministic between runs, and often lack single correct answers. Build evaluation frameworks that account for these characteristics, provide actionable feedback, catch regressions, and validate that context engineering choices achieve intended effects.

When to Activate

Activate this skill when:

Testing agent performance systematically
Validating context engineering choices
Measuring improvements over time
Catching regressions before deployment
Building quality gates for agent pipelines
Comparing different agent configurations
Evaluating production systems continuously

Do not activate this skill for adjacent work owned by other skills:

Designing the LLM judge itself, pairwise comparison, judge calibration, or bias mitigation: advanced-evaluation.
Designing autonomous control surfaces, novelty gates, rollback, or PR approval boundaries: harness-engineering.
Debugging a specific context failure mode before measuring it: context-degradation.