← 返回 Skills 市场

Axioma Skill Evaluator

Name: Axioma Skill Evaluator
Author: kofna3369

作者 Kofna3369 · GitHub ↗ · v2.2.0 · MIT-0

cross-platform ⚠ suspicious

总下载

当前安装

版本数

在 OpenClaw 中安装

/install axioma-skill-evaluator

功能描述

Advanced skill evaluation for OpenClaw agents. Use when: (1) evaluating a skill before publishing, (2) improving a skill based on evaluation results, (3) che...

使用说明 (SKILL.md)

AXIOMA SKILL EVALUATOR 🧙‍♂️

Advanced Skill Evaluation: Dual System (Automated + Manual)

Info	Value
Version	2.1.0 — 2026-05-07
Status	OPERATIONAL

1. PURPOSE AND SCOPE

Objective

Provide comprehensive skill evaluation using dual systems:

Axioma System (5 dimensions, 100 max) — colorful, fast
ISO 25010 System (25 criteria, 100 max) — international standard

When to Use

Trigger	Action
Before publishing a skill	Run both evaluations
Improving a skill	Get both automated + manual scores
Quality audit	Use 25-criteria rubric
Pre-publication check	Run all checks

2. BUNDLED TOOLS

evaluator.py (Axioma System)

# Run Axioma 5-dimension evaluation
python3 evaluator.py \x3Cskill-path> --verbose --improve

eval-skill.py (ISO 25010 System)

# Run automated ISO 25010 checks
python3 eval-skill.py \x3Cskill-path> --verbose

# JSON output
python3 eval-skill.py \x3Cskill-path> --json

3. AXIOMA EVALUATION SYSTEM

Quick Start

python3 evaluator.py \x3Cskill-path> --verbose --improve

5 Dimensions (100 max)

Dimension	Weight	Focus
Structure	20%	Header, sections, formatting, meta
Clarity	20%	Description, instructions, examples
Completeness	20%	Tools, prerequisites, errors, edge cases
Consistency	20%	Style, naming, integration
Functionality	20%	Commands work, expected results

Output Format

╔═══════════════════════════════════════════════════════════╗
║ 📊 SKILL EVALUATION REPORT — [Skill Name]              ║
║ Score: XX/100 [STATUS]                                  ║
╠═══════════════════════════════════════════════════════════╣
║ STRUCTURE:     XX/20 ████████████░░░░ XX%                ║
║ CLARITY:      XX/20 ████████████░░░░ XX%                ║
║ COMPLETENESS: XX/20 ████████████░░░░ XX%                ║
║ CONSISTENCY:  XX/20 ████████████░░░░ XX%                ║
║ FUNCTIONALITY: XX/20 ████████████░░░░ XX%                ║
╠═══════════════════════════════════════════════════════════╣
║ STATUS: ✅ APPROVED (score >= 70%)                      ║
╚═══════════════════════════════════════════════════════════╝

Thresholds

Score	Status	Action
90-100	🟢 EXCELLENT	Ready for production
70-89	🟡 GOOD	Publishable, minor notes
50-69	🟠 NEEDS_WORK	Fix before publishing
\x3C50	🔴 POOR	Major rework needed

4. ISO 25010 EVALUATION SYSTEM

Automated Checks (eval-skill.py)

Runs 13 automated checks:

File structure validation
Frontmatter YAML parsing
Description quality (65+ words, trigger contexts)
Script syntax validation
Credential scanning
Dependency audit

Target: 90%+ (12+/13 checks passed)

Manual Assessment (25 Criteria)

Category	Framework	Max	Criteria
1. Functional Suitability	ISO 25010	/12	Completeness, Correctness, Appropriateness
2. Reliability	ISO 25010	/12	Fault Tolerance, Error Reporting, Recoverability
3. Performance	ISO 25010	/8	Token Cost, Execution Efficiency
4. Usability (AI)	Shneiderman	/12	Learnability, Consistency, Feedback
5. Usability (Human)	Tognazzini	/8	Discoverability, Forgiveness
6. Security	ISO 25010	/12	Credentials, Input Validation, Data Safety
7. Maintainability	ISO 25010	/12	Modularity, Modifiability, Testability
8. Agent-Specific	Novel	/24	Trigger Precision, Progressive Disclosure, Composability
TOTAL		/100

5. COMPLETE EVALUATION WORKFLOW

1. AUTOMATED: python3 eval-skill.py \x3Cpath> --verbose
   → Target: 90%+ structural score
   ↓
2. AXIOMA: python3 evaluator.py \x3Cpath> --verbose --improve
   → Target: 70+ score
   ↓
3. MANUAL: Score 25 criteria rubric
   → Target: 80+ score
   ↓
4. FIX: Issues from all three sources
   ↓
5. RE-EVALUATE: Until all targets met
   ↓
6. PUBLISH: To ClawHub

6. ERROR HANDLING

Common Issues

Issue	Cause	Solution
No frontmatter	YAML not at start	Add `---` at start of SKILL.md
Poor description	Missing triggers	Add "Use when:" clauses
Empty directories	Unused folders	Remove or populate
Name mismatch	Directory ≠ frontmatter	Rename to match

Security Issues

Issue	Severity	Action
Hardcoded credentials	CRITICAL	Remove immediately
Missing input validation	HIGH	Add validation
No error handling	MEDIUM	Add try/catch blocks

7. EDGE CASES

Case	Input	Expected Output
Empty SKILL.md	Empty file	Error message, suggest template
Very long SKILL.md	>500 lines	Warning, recommend split
Missing description	No frontmatter	Fail with instructions
No scripts	No scripts/ dir	Pass, document as standalone

8. DEPENDENCIES

Dependency	Purpose	Required
Python 3.6+	Script execution	Yes
PyYAML	Frontmatter parsing	Optional

In Altum Per Quality. 🧙‍♂️ Axioma Skill Evaluator v2.1

安全使用建议

Review this skill before relying on it. Its purpose is legitimate and no exfiltration or destructive behavior is shown, but the bundled reports contain contradictory approval labels, and the Python tools read local skill files, scan for credentials, and use under-declared/environment-specific setup.

功能分析

Type: OpenClaw Skill Name: axioma-skill-evaluator Version: 2.2.0 The bundle is a legitimate utility for auditing and scoring OpenClaw skills based on structural and quality frameworks (Axioma and ISO 25010). The Python scripts (eval-skill.py, evaluator.py) perform static analysis, such as checking for hardcoded secrets and validating YAML frontmatter. While the scripts contain hardcoded local file paths (e.g., /media/ezekiel/Morgana/), there is no evidence of data exfiltration, unauthorized execution, or malicious prompt injection.

能力标签

cryptorequires-sensitive-credentials

能力评估

⚠ Purpose & Capability

The stated purpose is skill evaluation, but multiple bundled reports label below-threshold scores as approved, which could cause users or agents to over-trust the evaluator's results.

ℹ Instruction Scope

Instructions ask users to run bundled Python scripts on a chosen skill path; this is purpose-aligned but means the script will read local skill files.

ℹ Install Mechanism

The registry says there is no install spec and no required binaries, while SKILL.md requires Python 3.6+ and the code imports PyYAML. This looks like under-declared setup rather than hidden installation.

ℹ Credentials

The evaluator code contains environment-specific hardcoded paths under /media/ezekiel/Morgana/skills, which may not fit other users' systems and may create confusing report behavior.

ℹ Persistence & Privilege

The artifacts show report files are generated and bundled, but there is no evidence of background persistence, credential use for account access, or network exfiltration.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install axioma-skill-evaluator
安装完成后，直接呼叫该 Skill 的名称或使用 /axioma-skill-evaluator 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v2.2.0

v2.2: Fully impersonal, universal. Removed all Axioma Stellaris references. Dual evaluation system (Axioma 5-dim + ISO 25010) for any OpenClaw agent.

v2.1.0

v2.1: Dual evaluation system with Axioma (5-dim, 100max) + ISO 25010 (25 criteria, 100max). Added edge cases, clearer structure, improved commands documentation.

v2.0.0

Merged Axioma evaluator + terwox/skill-evaluator. Includes eval-skill.py automated checks, 5-dimension cluster evaluation, 25-criteria ISO 25010 rubric, colorful terminal output. Complete skill quality solution.

v1.0.0

Bilingual: SKILL.md (French) + SKILL_EN.md (English). Cluster Axioma Stellaris skill evaluation tool.

元数据

Slug axioma-skill-evaluator

版本 2.2.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 4

常见问题