eval

Evaluate and rank agent results by metric or LLM judge for an AgentHub session. Use when the user runs /hub:eval or asks to score, compare, or pick a winner among completed AgentHub agents.

Quality

76%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

—

The risk profile of this skill

Fix and improve this skill with Tessl

tessl review fix ./engineering/agenthub/skills/eval/SKILL.md

Quality

Content

68%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 compact, action-oriented skill body with concrete commands and a clear mode-based structure. Its main weakness is workflow clarity: a batch evaluation across multiple agents lacks explicit validation/verification checkpoints and error-recovery feedback loops.

Suggestions

Add a validation step after running the eval command — e.g. check the ranker exit code and handle failures before declaring a winner — to introduce a feedback loop and lift workflow_clarity above the batch-operation cap of 3.

Confirm the referenced scripts exist as bundle files (scripts/result_ranker.py, scripts/session_manager.py) or note that they are provided by the AgentHub runtime, so the in-body references are not dangling.

Resolve or document the template placeholders ({session-id}, {eval_cmd}, {metric}, {direction}) so the example commands read as copy-paste ready rather than schematic.

Dimension	Reasoning	Score
Conciseness	The body is efficient and assumes Claude's competence, using compact command blocks and example output tables rather than explanatory prose; a few illustrative output blocks could be trimmed, so it falls just short of a 5.	4 / 5
Actionability	Provides concrete, mostly executable commands ('python {skill_path}/scripts/result_ranker.py ...', 'git diff {base_branch}...{agent_branch}') and explicit ranking criteria; minor gaps come from unresolved template placeholders like {session-id} and {eval_cmd}.	4 / 5
Workflow Clarity	The three modes and 'After Eval' steps are sequenced, but this is a batch operation over multiple agents with no validation that the eval command succeeded and no error-recovery loop; per the rubric cap, a destructive/batch workflow without validation cannot exceed 3.	3 / 5
Progressive Disclosure	Content is organized into clearly labeled sections (Usage, What It Does with sub-modes, After Eval) in a single well-structured file with no nested references; at >50 lines it does not qualify for the simple-skill 5, and the referenced script paths are not present as bundle files, so it stays at 4.	4 / 5
	Total	15 / 20 Passed

Description

83%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 well-crafted description that clearly states both capability and trigger conditions in third-person imperative voice, with concrete actions and natural trigger terms. It is held back from perfection only by slightly less-than-exhaustive action and keyword enumeration.

Dimension	Reasoning	Score
Specificity	Lists several concrete actions — 'Evaluate and rank agent results by metric or LLM judge' plus 'score, compare, or pick a winner' — with only minor gaps in coverage; not a full 5 because it stops short of enumerating the ranking modes themselves.	4 / 5
Completeness	Explicitly answers both what ('Evaluate and rank agent results by metric or LLM judge for an AgentHub session') and when ('Use when the user runs /hub:eval or asks to score, compare, or pick a winner among completed AgentHub agents') with concrete trigger phrases.	5 / 5
Trigger Term Quality	'asks to score, compare, or pick a winner among completed AgentHub agents' captures natural user phrasings well; a few synonyms (e.g. 'grade', 'review results') are missing, keeping it below 5.	4 / 5
Distinctiveness Conflict Risk	The 'for an AgentHub session' and '/hub:eval' framing carves a clear niche with distinct triggers; minor overlap risk with generic evaluation skills keeps it just below 5.	4 / 5
	Total	17 / 20 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: alirezarezvani/claude-skills
Path: engineering/agenthub/skills/eval/SKILL.md
Commit: 19392f7

Reviewed: ~7 hours 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.