← ClaudeAtlas

security-review-trainerlisted

Progressive security review challenges -- intentional vulnerabilities embedded in clean code, scored findings, and increasing subtlety. Use when building security review skills, practicing vulnerability identification, developing severity judgment, or training to detect subtle security flaws in progressively harder code samples.
michaelalber/ai-toolkit · ★ 1 · AI & Automation · score 77
Install: claude install-skill michaelalber/ai-toolkit
# Security Review Trainer > "The art of security is not in finding what is obviously broken, but in recognizing what > should not be trusted in code that appears to work correctly." > -- Gary McGraw, Software Security: Building Security In > "Defenders think in lists. Attackers think in graphs." > -- John Lambert, Microsoft Threat Intelligence ## Core Philosophy Security review is a skill that atrophies without practice. Most developers can spot obviously dangerous patterns like unsanitized user input passed to a shell command, but miss subtle IDOR vulnerabilities, timing side-channels, or deserialization traps hidden in otherwise clean code. This skill generates progressively harder security challenges where you must find intentionally planted vulnerabilities — building pattern recognition that makes security review instinctive rather than checklist-dependent. **Why precision matters as much as recall:** A reviewer who flags 15 items in every PR, most non-issues, trains their team to ignore security comments. False positives are not harmless — they create an illusion of coverage while eroding trust in security feedback. This trainer scores both what you found AND what you incorrectly flagged. **The CACR loop:** Challenge → Attempt → Compare → Reflect. Each cycle presents realistic code with intentionally planted vulnerabilities at a calibrated difficulty level. **What this skill does NOT do:** This skill teaches defensive security review — finding vulnerabilities befo