skill-tester
by alirezarezvaniskill-tester is a QA meta-skill for validating, testing, and scoring claude-skills directories. Use its Python stdlib tools to check structure, script syntax/runtime, quality grades, tier fit, and optional security posture for Skill Testing workflows.
This skill scores 82/100, which makes it a solid listing candidate for directory users who work inside the claude-skills ecosystem. It provides concrete validation and scoring scripts, runnable examples, references, and expected output, so an agent can do more than simply follow a generic QA prompt. The main caveat is that its rules are ecosystem-specific and partly legacy-oriented, so users should install it when they want skill QA for this repository style rather than universal skill validation.
- Strong triggerability: the description clearly says to use it when authoring, auditing, promoting, pre-commit gating, or CI-testing skills.
- Operationally useful tooling: includes four Python scripts for structure validation, script testing, quality scoring, and security scoring, with JSON-capable outputs and sample reports.
- Good supporting context: references include a quality scoring rubric, skill structure specification, and tier requirements matrix, reducing agent guesswork.
- Focused on the claude-skills repository conventions, including legacy tier and line-count rules, so results may not generalize cleanly to other skill formats.
- No install command is provided in SKILL.md, and README examples use paths that may require users to understand whether they are running from the skill directory or repo root.
Overview of skill-tester skill
What skill-tester is for
skill-tester is a QA meta-skill for validating, testing, and scoring skills in the claude-skills ecosystem. It is useful when you are building a new skill, auditing an existing skill before publishing, checking whether a skill meets a BASIC/STANDARD/POWERFUL tier target, or adding repeatable quality gates to a repository workflow.
Best-fit users and jobs
The skill-tester skill is best for skill authors, maintainers, reviewers, and engineering teams who need more than a subjective “looks good” review. Its core job is to turn a skill directory into actionable evidence: structure compliance, script health, quality grade, tier recommendation, and optionally security posture.
What makes it different from a generic review prompt
Instead of relying only on free-form critique, skill-tester includes Python standard-library tools that inspect files and produce repeatable results. The main tools are skill_validator.py for structure, script_tester.py for Python scripts, quality_scorer.py for weighted scoring, and security_scorer.py for security checks or --include-security scoring.
Important adoption caveat
The repository notes that some tier line-count minimums reflect legacy skills. For new skill authoring, do not pad SKILL.md just to satisfy a tier target. Use skill-tester as a QA and audit tool, then reconcile its findings with the current authoring standard used by your project.
How to Use skill-tester skill
skill-tester install and repository context
Install the skill from the source repository with:
npx skills add alirezarezvani/claude-skills --skill skill-tester
The upstream skill lives at:
engineering/skills/skill-tester
After install, read SKILL.md first for the exact workflow, then README.md for runnable examples. If you need to understand scoring decisions, inspect references/quality-scoring-rubric.md, references/skill-structure-specification.md, and references/tier-requirements-matrix.md. The sample output in expected_outputs/sample_validation_report.json is useful for seeing the shape of machine-readable results.
Run the core checks from the repo root
The scripts are designed to be run from the repository root with full paths. Typical skill-tester usage looks like this:
python3 engineering/skills/skill-tester/scripts/skill_validator.py engineering/my-skill
python3 engineering/skills/skill-tester/scripts/script_tester.py engineering/my-skill --timeout 60 --json
python3 engineering/skills/skill-tester/scripts/quality_scorer.py engineering/my-skill --detailed --json
python3 engineering/skills/skill-tester/scripts/quality_scorer.py engineering/my-skill --detailed --include-security --json
Use JSON output when you want CI gates, pre-commit checks, dashboards, or comparison over time. Use the human-readable mode when you are interactively fixing a skill.
Give the agent a complete testing prompt
A weak prompt is: “Test my skill.” A stronger prompt for the skill-tester skill is:
Use skill-tester to audit `engineering/skills/my-skill`.
Run structure validation, script testing with a 60 second timeout, detailed quality scoring, and security-inclusive scoring.
Return a prioritized fix list grouped by: blocking failures, quality improvements, security concerns, and optional polish.
Do not recommend padding documentation purely to satisfy legacy line-count thresholds.
This prompt works better because it gives the skill path, the checks to run, the timeout, the desired output format, and a policy constraint for interpreting tier rules.
Practical workflow for better results
Start with skill_validator.py because missing files and malformed frontmatter can make later scoring noisy. Then run script_tester.py to catch syntax, import, runtime, argparse, and output-format problems. Use quality_scorer.py --detailed after the obvious failures are fixed, because its improvement roadmap is more useful once the skill is structurally valid. Add --include-security when scripts touch files, subprocesses, user input, credentials, network calls, or generated code.
skill-tester skill FAQ
Is skill-tester only for this repository?
It is written for the claude-skills ecosystem and assumes that style of skill directory, including SKILL.md, README.md, scripts/, optional assets/, references/, and expected_outputs/. You can adapt the logic elsewhere, but the scoring and tier assumptions are not universal software-quality standards.
Can beginners use skill-tester?
Yes, if they can run Python scripts from a terminal and understand file paths. The tools use the Python standard library, which lowers setup friction. Beginners should start with one command at a time and compare their results with assets/sample-skill and expected_outputs/sample_validation_report.json.
How is skill-tester different from unit tests?
Unit tests usually verify application behavior. skill-tester checks whether a skill package is well-formed, documented, script-compatible, scoreable, and safe enough for the intended tier. It complements tests; it does not replace domain-specific test cases for the skill’s actual business logic.
When should I not use skill-tester?
Do not use it as the only judge for non-claude-skills projects, prose-only prompts, or skills that intentionally do not follow this directory model. Also avoid treating the letter grade as a product-quality guarantee. A skill can score well structurally while still being strategically unnecessary or poorly suited to your users.
How to Improve skill-tester skill
Improve skill-tester inputs before rerunning
The fastest way to get better skill-tester results is to make the target skill easier to inspect. Ensure SKILL.md has clear frontmatter, README.md has runnable examples, scripts use argparse, scripts include a main guard, and expected outputs show real result shapes. If your skill has dependencies or external services, document setup and safe failure behavior before scoring.
Fix common failure modes in order
Treat failures in this order: missing required files, broken Python syntax/imports, scripts that hang or require hidden state, unclear usage examples, weak expected outputs, then tier or scoring polish. This order matters because high-level quality recommendations are less reliable when the tester cannot execute or parse the basic structure.
Use scoring as an iteration loop
After the first skill-tester run, ask for a compact remediation plan rather than a full rewrite:
Based on the latest skill-tester JSON reports, identify the smallest set of changes that would raise this skill from B to A-. Separate changes that affect actual usability from changes that only affect rubric compliance.
This keeps the work focused on user-visible quality: clearer invocation, safer scripts, better examples, and more complete references.
Extend cautiously for your team’s standards
If your organization has stricter requirements, adapt the scripts rather than relying on manual review. Good extensions include custom minimum scores for CI, security gates for file or subprocess use, additional expected-output checks, and project-specific documentation rules. Keep changes transparent by updating the relevant reference rubric so future reviewers know why the skill-tester result changed.