agentleFS
Sign inStart free
m00kkm00kk/agent-skills-playbook/skills/evaluate-agent-quality/SKILL.md

evaluate-agent-quality

Skill0 starsChanged 6 months agoMITView on GitHub

Creates regression suites and golden-path evaluations for AI agents: task success, tool correctness, citation faithfulness, and safety refusals. Use when testing agents, CI eval gates, or benchmarking prompt or tool changes.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use separate model or fixed rubric prompt

    line 47
  2. Distinctive:

    Always keep deterministic checks (tool called, file exists)

    line 49
  3. Distinctive:

    Run subset (smoke 5 cases) on every PR

    line 53

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Evaluate Agent Quality

Eval types

TypeMeasures
Task successDid the agent complete the user goal?
Tool accuracyRight tool, right args, right order
RAG faithfulnessAnswer supported by retrieved chunks
SafetyRefusal on harmful or out-of-scope requests
RegressionNo worse than baseline on golden set

Golden dataset format

{
  "id": "rag-001",
  "input": "What is our refund policy for EU customers?",
  "expected_tools": ["search_knowledge"],
  "expected_citations": ["policy-eu-2024.md"],
  "rubrics": ["mentions 14-day window", "cites EU doc"]
}

Store under evals/golden/ in the project (not in skill repo).

Workflow

- [ ] Collect 20–50 real user tasks (anonymized)
- [ ] Label expected outcome + forbidden behaviors
- [ ] Automate runner: invoke agent, capture trace
- [ ] Score with rules + optional LLM-as-judge
- [ ] Block deploy if success rate drops >5% vs baseline

LLM-as-judge cautions

  • Use separate model or fixed rubric prompt
  • Judges fail on subjective tone — prefer binary checks
  • Always keep deterministic checks (tool called, file exists)

CI integration

  • Run subset (smoke 5 cases) on every PR
  • Full suite nightly
  • See github-ci-agents for Actions wiring

Metrics to track over time

  • pass@1, pass@3
  • average steps per task
  • cost per successful task

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free