scienceworld-object-selector

Use when the agent needs to choose a specific object from multiple candidates in the environment based on task criteria such as object type (non-living thing, electrical component, container), properties, or category. This skill surveys visible objects with "look around", evaluates each against the task requirements, selects the most suitable candidate, and signals intent with "focus on [OBJECT]".

Quality

83%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

78%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 lean, highly actionable single-purpose skill with concrete commands and two complete worked examples, and a clear sequenced workflow. Its main weakness is progressive disclosure: the provided reference file object_properties.md is never mentioned in the body, so its useful classification content is undiscoverable.

Suggestions

Reference references/object_properties.md from the body (e.g., in the Evaluate candidates step or a 'References' section) so the classification tables and selection heuristics are discoverable.

Promote the 'examine [OBJECT]' verification into the workflow as an explicit conditional checkpoint step rather than only a key principle.

Trim the Purpose section and the second example transcript, which restate information already conveyed elsewhere.

Dimension	Reasoning	Score
Conciseness	The body is efficient with no explanations of concepts Claude already knows, but the Purpose and When to Use sections partially restate the frontmatter description and the second example transcript could be trimmed. Not 5 because these minor redundancies could be tightened.	4 / 5
Actionability	Fully executable environment commands ('look around', 'focus on black wire', 'move black wire to purple box', 'examine [OBJECT]') plus two complete worked transcripts covering the common cases (non-living thing, electrical component). Copy-paste ready with observed outputs.	5 / 5
Workflow Clarity	A clear five-step numbered sequence (survey, evaluate, select, signal, execute) with concrete exclusion rules, but verification is only mentioned as a Key Principle ('Use examine [OBJECT]... if uncertain') rather than an explicit in-workflow checkpoint. Not 5 because there is no in-line validation or feedback loop.	4 / 5
Progressive Disclosure	The body itself is well-sectioned, but the bundle provides references/object_properties.md (classification tables and selection heuristics directly relevant to the evaluate step) that is never referenced or linked from SKILL.md — an orphaned reference that is not signaled at all, so navigation to the deeper material is broken. Not 4 because per the bundle-structure guideline the unreferenced file is a real organization gap, not a minor one.	3 / 5
	Total	16 / 20 Passed

Description

88%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 explicitly covers both what the skill does and when to use it, with concrete commands and task-criteria examples. The main gaps are a few missing natural trigger synonyms and the unnamed target environment, which slightly raises conflict risk.

Suggestions

Name the ScienceWorld environment explicitly to eliminate overlap risk with generic object-selection skills.

Add common trigger variations such as 'find' or 'identify' an object to broaden natural keyword coverage.

Dimension	Reasoning	Score
Specificity	The description lists four concrete actions with actual command names ("surveys visible objects with 'look around'", "evaluates each against the task requirements", "selects the most suitable candidate", "signals intent with 'focus on [OBJECT]'"), giving comprehensive coverage of the skill's workflow.	5 / 5
Completeness	It explicitly answers both: what ("surveys visible objects... evaluates... selects... signals intent with 'focus on [OBJECT]'") and when ("Use when the agent needs to choose a specific object from multiple candidates... based on task criteria"), with concrete trigger criteria. Not 4 because neither the what nor the when is weakly specified.	5 / 5
Trigger Term Quality	It includes natural task vocabulary such as "choose a specific object from multiple candidates", "non-living thing, electrical component, container", and "properties, or category", but misses common variations users/tasks would phrase like "find" or "identify" an object.	4 / 5
Distinctiveness Conflict Risk	The distinctive environment commands ('look around', 'focus on [OBJECT]') carve out a clear niche, but the environment itself (ScienceWorld) is never named, leaving minor overlap risk with generic object-selection or environment-navigation skills.	4 / 5
	Total	18 / 20 Passed

Validation

100%

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 — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: zjunlp/SkillNet
Path: experiments/src/skills/scienceworld/scienceworld-object-selector/SKILL.md
Commit: c490239

Reviewed: 3 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.