agentv-eval-review

Eval Review

Overview

Lint and review AgentV eval YAML files for structural issues, schema compliance, and quality problems. Apply this checklist deterministically first, then layer LLM judgment for semantic issues a checklist cannot catch.

Process

Step 1: Structural checklist

Walk every target eval file and report violations grouped by severity (error > warning > info). For each finding, include the file path and a concrete fix.

• File extension is .eval.yaml (error if not).
• description field is present at the top level (error if missing).
• Each entry under tests has id, input, and at least one of criteria / expected_output / assert (error if missing).
• File-typed inputs (type: file) use a leading / in their path (error if relative).
• Tests have an assert block — flag tests that rely solely on expected_output (warning).
• Flag criteria that duplicates assertion strings when assertions already express the grading contract (warning — remove the duplicate criteria).
• Prefer plain assertion strings over multiple named type: llm-rubric blocks when the default LLM rubric grader can evaluate the checks (info unless custom prompts or grader targets are present).
• Detect expected_output prose patterns like "The agent should..." or "The output is..." (warning — expected_output should be a golden/reference answer; scoring rules belong in assertions or, for implicit-grader cases, criteria).
• For historical or repo-state evals, verify the relevant repo is pinned under workspace.repos[].commit; a SHA mentioned only in prompt prose or metadata is not an operational checkout (warning).
• Identical file inputs repeated across multiple tests in the same eval should be hoisted to a top-level input (info).
• Eval files in the same directory should share a common id prefix (info — flag drift).

Step 2: Semantic review (LLM judgment)

The structural checklist catches mechanical issues but cannot assess:

• Factual accuracy — Do tool/command names in expected_output match what the skill documents?
• Coverage gaps — Are important edge cases missing?
• Assertion discriminability — Would assertions pass for both good and bad output?
• Cross-file consistency — Do output filenames match across evals and skills?