advanced-evaluation

This skill should be used when the user asks to "implement LLM-as-judge", "compare model outputs", "create evaluation rubrics", "mitigate evaluation bias", or mentions direct scoring, pairwise comparison, position bias, evaluation pipelines, or automated quality assessment.

Quality

73%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./skills/advanced-evaluation/SKILL.md

Quality

Content

65%Weight 40%Scale 1-3

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

The body is highly actionable with concrete templates, protocols, and JSON examples, but it is verbose (teaching known concepts), its main pipeline lacks explicit validation checkpoints, and it is a monolith with internal references that resolve to no actual files.

Suggestions

Cut conceptual teaching Claude already knows — collapse the bias-definitions and metric-selection table into terse mitigation/selection rules — to tighten conciseness toward a 3.

Add explicit validation checkpoints to the Evaluation Pipeline (e.g., validate rubric weights sum to 1.0 before scoring, retry low-confidence judgments) so the primary workflow has feedback loops.

Move the full prompt templates, JSON examples, and the metric/rubric reference material into real files under references/ and link to them one level deep, replacing the dangling 'Internal reference' name list.

Dimension	Reasoning	Score
Conciseness	The ~450-line body spends substantial tokens teaching concepts Claude already knows (definitions of position/length/self-enhancement/verbosity/authority bias and a metric table explaining Recall, F1, Cohen's kappa, Spearman's rho), so it is mostly efficient but includes unnecessary explanation that could be tightened.	2 / 3
Actionability	Provides concrete, copy-paste-ready prompt templates for direct scoring and pairwise comparison, a numbered position-swap protocol, strictness-calibration guidance, and complete JSON output examples rather than vague or pseudocode direction.	3 / 3
Workflow Clarity	Sequences are present (the position-swap protocol and pipeline stages are ordered), but the centerpiece Evaluation Pipeline diagram has no explicit validate/retry checkpoints between stages, leaving validation implicit as in the 'steps listed but validation gaps' anchor.	2 / 3
Progressive Disclosure	The skill is a 450-line monolith with all content inline and the 'Internal reference' entries (LLM-as-Judge Implementation Patterns, Bias Mitigation Techniques, Metric Selection Guide) are dangling names with no corresponding files in references/, so structure exists but content that should be split is inline and references are not properly externalized.	2 / 3
	Total	9 / 12 Passed

Description

82%Weight 40%Scale 1-3

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

The description has strong, natural trigger terms and a distinctive niche, but it frames everything as a trigger clause without an explicit capability statement, leaving the 'what does this do' half of completeness only implied.

Suggestions

Lead with an explicit capability statement (e.g., 'Builds LLM-as-judge evaluation systems: designs rubrics, runs direct and pairwise scoring, and mitigates evaluation bias.') before the 'Use when...' trigger clause to fully answer both what and when.

Tighten 'compare model outputs' and 'automated quality assessment', which are relatively broad, with more distinctive phrasing to reduce overlap with a general evaluation skill.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions ("implement LLM-as-judge", "compare model outputs", "create evaluation rubrics", "mitigate evaluation bias") rather than vague language, matching the 'lists multiple specific concrete actions' anchor.	3 / 3
Completeness	The 'when' is explicit ('This skill should be used when the user asks to...') but the 'what' is only implied through the quoted trigger actions; there is no separate explicit capability statement, so it does not clearly answer both what and when as required for a 3.	2 / 3
Trigger Term Quality	Includes natural quoted phrases a user would say ("implement LLM-as-judge", "compare model outputs", "create evaluation rubrics") plus mentions of direct scoring, pairwise comparison, and position bias, giving good coverage of natural trigger terms.	3 / 3
Distinctiveness Conflict Risk	The LLM-as-judge / bias-mitigation niche with distinctive jargon (pairwise comparison, position bias, evaluation rubrics) is unlikely to trigger for unrelated skills, fitting the 'clear niche with distinct triggers' anchor.	3 / 3
	Total	11 / 12 Passed

Validation

93%

Warnings & errors only

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

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: sickn33/antigravity-awesome-skills
Path: skills/advanced-evaluation/SKILL.md
Commit: 07da495

Reviewed: 9 days ago

Table of Contents

Discovery Implementation Validation

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.