CtrlK
BlogDocsLog inGet started
Tessl Logo

write-judge-prompt

Design LLM-as-Judge evaluators for subjective criteria that code-based checks cannot handle. Use when a failure mode requires interpretation (tone, faithfulness, relevance, completeness). Do NOT use when the failure mode can be checked with code (regex, schema validation, execution tests). Do NOT use when you need to validate or calibrate the judge — use validate-evaluator instead.

59

Quality

68%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./skills/write-judge-prompt/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

61%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A well-structured, actionable guide to designing binary judge prompts with concrete templates and rules. The main weaknesses are verbosity in the illustrative examples and validation being implicit rather than an explicit workflow checkpoint.

Suggestions

Tighten the few-shot example block (shorten the full email texts and the Likert-scales rationale) to lift conciseness toward 4-5.

Promote validation into the main workflow as an explicit final step (e.g., '5. Validate: run validate-evaluator against human labels before trusting the judge') instead of only mentioning it under Anti-Patterns.

Add one fully assembled end-to-end judge prompt combining all four components, or move the long example block into a references/ file, to push actionability and progressive disclosure higher.

DimensionReasoningScore

Conciseness

Mostly on-point methodology, but the full inlined email few-shot examples and the lengthy Likert-scales rationale are padding that could be trimmed without losing the method, fitting 'mostly efficient but could be tightened' rather than the lean 4-5 anchors.

3 / 5

Actionability

Provides concrete per-component templates, an explicit JSON output schema, and specific rules (2-4 examples, require a borderline case, critique before verdict); the only gap is no single assembled end-to-end prompt combining all four components, so it stops short of fully copy-paste ready.

4 / 5

Workflow Clarity

A prerequisite gate and a numbered four-component sequence are present, but validation is relegated to the Anti-Patterns section rather than appearing as an explicit checkpoint in the main flow, matching 'sequence present but checkpoints implicit'.

3 / 5

Progressive Disclosure

No bundle files exist and none are needed; the single file is well-organized into clear sections, but at ~130 lines the long inline example block could be split into a reference file, so it is not a clean 5.

4 / 5

Total

14

/

20

Passed

Description

75%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong description that clearly answers both what and when, with explicit do-not-use boundaries that sharpen distinctiveness. Specificity and natural trigger-term coverage are good but not comprehensive, capping those dimensions at 3.

DimensionReasoningScore

Specificity

Names the domain ('LLM-as-Judge evaluators for subjective criteria') and one concrete action ('Design'), but does not list multiple specific actions, matching the 'domain and 1-2 concrete actions' anchor rather than the several-actions anchors at 4-5.

3 / 5

Completeness

Explicitly states both what ('Design LLM-as-Judge evaluators...') and when ('Use when a failure mode requires interpretation...') with concrete trigger phrases, and adds do-not-use boundaries, matching the top anchor.

5 / 5

Trigger Term Quality

The parenthetical 'tone, faithfulness, relevance, completeness' supplies natural terms, but the dominant framing ('failure mode', 'LLM-as-Judge', 'code-based checks') is more technical than the natural phrases a user would say, leaving common variations uncovered.

3 / 5

Distinctiveness Conflict Risk

It carves a clear niche and explicitly disambiguates from a sibling skill ('use validate-evaluator instead'), giving minimal conflict risk.

5 / 5

Total

16

/

20

Passed

Validation

100%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository
hamelsmu/evals-skills
Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.