AI Security Assessment
Scan prompts and AI systems for injection, jailbreak and poisoning risks, mapped to MITRE ATLAS.
- Repository
- alirezarezvani/claude-skills
- Stars
- 29,149
- Licence
- MIT
- Files
- 3 files
What it does
This skill teaches an AI agent to assess AI and LLM-based systems for weaknesses such as prompt injection, jailbreak attempts, training-data leakage (model inversion), data poisoning and agent tool abuse. It bundles a Python scanner that checks test prompts against regex signatures for known attack phrasing, labels findings with MITRE ATLAS technique IDs, and returns a risk exit code. The skill itself is methodology: it tells the agent how to structure the assessment and recommend guardrails. The scanner is static pattern matching on text; it does not test a live model and does not prove a system is secure.
Great for
- Engineers adding an LLM feature who want a first-pass check of prompts and guardrails before release
- Security teams running a documented, authorized review of an AI assistant or agent
- Teams building RAG or web-browsing agents who need to understand indirect injection risk
Skip it if
- Certifying an AI system as safe: the regex scanner misses many attacks and reports only signature matches, so a clean result is not evidence of security
- Testing systems you do not own or have written permission to test: the skill explicitly requires authorization for gray-box and white-box testing
- General application or infrastructure security work, which the skill itself says belongs to other skills
Try asking
- “Scan my chatbot's built-in seed prompts for prompt injection signatures using black-box access”
- “Check this prompts.json test file for jailbreak patterns and map findings to MITRE ATLAS”
- “Assess our agent's tool-call exposure and list which injection signatures are critical”
- “Explain which ATLAS techniques apply to our RAG pipeline that reads web pages”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 3 original files, the MIT licence and an attribution note.
ai-security.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip ai-security.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip ai-security.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Use --list-patterns first to see which signatures exist, so you understand what the scanner can and cannot detect
- Supply your own realistic attack prompts via --test-file; the built-in seeds are only examples
- Treat the scanner output as a starting checklist and pair it with human review and guardrail testing
What's inside
- references/atlas-coverage.md
- scripts/ai_threat_scanner.py
- SKILL.md
Redistributed unmodified from alirezarezvani/claude-skills under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.