Name: ai-security
Rating: 67.2 (1 reviews)
Author: alirezarezvani

ai-security

Use when assessing AI/ML systems for prompt injection, jailbreak vulnerabilities, model inversion risk, data poisoning exposure, or agent tool abuse. Covers MITRE ATLAS technique mapping, injection signature detection, and adversarial robustness scoring.

Quality

81%

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

The canonical home for this skill is ai-security in alirezarezvani/claude-skills

Quality

Content

71%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 well-organized, actionable body with concrete commands, clear workflows, and useful decision checkpoints. Its main weakness is progressive disclosure: it references bundle files (scripts/ai_threat_scanner.py, references/atlas-coverage.md) that are not present, and inlines content that would benefit from being split into reference files.

Suggestions

Ship the referenced bundle files — create scripts/ai_threat_scanner.py and references/atlas-coverage.md — or remove the references so the skill is self-contained and commands are actually executable.

Move the large signature and ATLAS coverage tables into reference files (e.g. references/injection-signatures.md) and keep SKILL.md as an overview pointing to them, deepening the progressive-disclosure structure.

Trim definitional prose (e.g. 'Prompt injection occurs when...', 'Model inversion attacks reconstruct...') that restates concepts Claude already knows, and add per-iteration validation to the batch workflow loop.

Dimension	Reasoning	Score
Conciseness	Mostly dense and efficient — tables, concrete commands, and exit codes earn their place — but a few definitional sentences ('Prompt injection occurs when adversarial input overrides...', 'Model inversion attacks reconstruct training data...') explain concepts Claude already knows and could be trimmed.	4 / 5
Actionability	Provides concrete, copy-paste-ready commands with specific flags, a test-file JSON format, and exit-code semantics; however the referenced `scripts/ai_threat_scanner.py` is not bundled, so the commands are not actually runnable as-is — a minor gap below fully executable.	4 / 5
Workflow Clarity	Three clearly sequenced workflows with explicit decision checkpoints (exit code 2 = block deployment, CI/CD risk gate) and phased structure; the batch `for target` loop lacks per-iteration validation, a minor gap keeping it below 5.	4 / 5
Progressive Disclosure	Good in-body structure (TOC, headed sections, one signaled reference to `references/atlas-coverage.md`), but neither that reference nor `scripts/ai_threat_scanner.py` exist in the bundle, and substantial reference-style tables (signatures, ATLAS coverage) are inlined rather than split out.	3 / 5
	Total	15 / 20 Passed

Description

92%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, well-scoped description that explicitly states both what the skill does and when to use it, with concrete natural-language trigger terms and a distinct AI/ML security niche. Only minor gap is a few missing keyword synonyms keeping trigger_term_quality just below maximum.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — 'assessing...for prompt injection, jailbreak vulnerabilities, model inversion risk, data poisoning exposure, or agent tool abuse' plus 'MITRE ATLAS technique mapping, injection signature detection, and adversarial robustness scoring' — giving comprehensive coverage of the skill's capabilities.	5 / 5
Completeness	Explicitly answers both 'what' ('Covers MITRE ATLAS technique mapping, injection signature detection, and adversarial robustness scoring') and 'when' ('Use when assessing AI/ML systems for...') with concrete trigger phrases, matching the 5 anchor.	5 / 5
Trigger Term Quality	Strong natural keywords users would actually say ('prompt injection', 'jailbreak', 'model inversion', 'data poisoning', 'agent tool abuse'), but a few common variants/synonyms are absent, so it sits just below the comprehensive 5 anchor.	4 / 5
Distinctiveness Conflict Risk	Clear niche (AI/ML system security specifically) with distinct triggers; the body further disambiguates from security-pen-testing, red-team, and threat-detection, giving minimal conflict risk.	5 / 5
	Total	19 / 20 Passed

Validation

87%

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 — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
relative_links	Relative link issues: 4 suspicious	Warning
referenced_paths_exist	Referenced path issues: 10 missing	Warning

	Total	14 / 16 Passed

Repository: alirezarezvani/claude-skills
Path: .gemini/skills/ai-security/SKILL.md
Commit: 19392f7

Reviewed: ~1 hour 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.