Evaluate a research report against its stated topic across six dimensions (relevance, completeness, source authority, depth balance, consistency, decision quality), assigning a conceptual quality level and issues to each.
SkillScore41Maintained
Automatic
◆Research pack
Last verified 8/12/2026
What it changes
Passed 9 of 22 test cases with the skill, and 5 of 22 without it.
Without skill23%
With skill41%
Checks it passed that the model alone missed
Evaluates depth balance.
Evaluates internal consistency.
Evaluates coverage completeness.
and 16 more in the full report
8 fixed1 passed either way9 still failing4 broke
Pass-rate lift
+18 pp
Tokens
+27%
Agent turns
0%
Response time
+17%
Model gemini-3.6-flashMeasured 8/12/2026v19 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 8 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 8
Prompt
Review our market readiness report for fault-tolerant quantum computing systems. The draft lists expert interviews and patent landscape data. Judge the report strictly against its explicit contents rather than unstated potential.
What was checked, without and with the skill
Checked
Without skill
With skill
Evaluates internal consistency within the document.
Missed
Passed
Evaluates decision quality for executive action.
Missed
Passed
Bases evaluation strictly on content present in the document.
Missed
Passed
case-08 · gemini-3.6-flash · 8/12/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md13 linessingle file
Research Report Evaluator
You are a rigorous research reviewer. First note what the report claims to cover
and what sources/methodology it states. Then evaluate it across these six
dimensions: content relevance, coverage completeness, source authority, depth
balance, internal consistency, and decision quality.
For each dimension assign a conceptual level — EXCELLENT / ADEQUATE / WEAK /
FAILED — with 1-2 specific issues and one actionable recommendation. Then give an
overall verdict (e.g. publish / revise / reject) with a one-paragraph rationale.
Judge only what the report actually contains. Reward grounded, well-sourced,
decision-useful analysis; penalize unsupported claims and missing tradeoffs.