skills.rest
MoofonLi
MoofonLiCOMMUNITY
Explore all skills by this author →
GitHub

security-audit

Audits codebases for exploitable vulnerabilities using parallel sub-agent hunting and adversarial validation.

2|Updated Jul 11, 2026
Search Tags:#code review#vulnerability#penetration testing#security audit#prompt injection#exploit#findings report
One-click install
npx skills add https://github.com/MoofonLi/dev-ready --skill security-audit-moofonli
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: security-audit
Source: https://github.com/MoofonLi/dev-ready/tree/main/src/dev_ready/templates/claude/skills/security-audit
Command: npx skills add https://github.com/MoofonLi/dev-ready --skill security-audit-moofonli

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Manual security reviews miss subtle bugs like business logic flaws, chained attacks, and second-order injection that automated scanners cannot find, while unstructured reviews produce false positives and unverifiable claims. ## Core Features & Use Cases - Six-Phase Audit Workflow: Reconnaissance, parallel vulnerability hunting, adversarial validation, reporting, schema-checked structured output, and independent verification of every finding. - Domain-Specific Attack Classes: Specialized hunting guides for web apps, AI/LLM agents, memory-unsafe native code, HTTP protocol and auth layers, and client-side browser surfaces. - Machine-Readable Findings: Produces findings.json validated against a strict JSON schema alongside human-readable REPORT.md and FINDINGS-DETAIL.md files. - Use Case: Point the skill at a FastAPI backend repository and receive a report of confirmed, severity-rated vulnerabilities with concrete attack payloads, file:line traces, and remediation code. ## Quick Start Ask your coding agent to run a security audit on the current repository and write the findings report to an output directory.

Frequently Asked Questions about security-audit

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a security audit on my codebase with an AI agent?▼

Invoke the security-audit skill with the target repository path and an output directory. It runs six phases: reconnaissance, parallel vulnerability hunting, adversarial validation, report writing, schema-checked JSON output, and independent verification of each finding.

What types of vulnerabilities does a security audit skill find?▼

It covers injection, access control, business logic flaws, chained attacks, cryptography misuse, and feature abuse. Domain files add prompt injection for AI/LLM targets, memory safety for native code, request smuggling and JWT defects for HTTP/auth layers, and DOM XSS for client-side code.

Does the security audit work on AI and LLM applications?▼

Yes, a dedicated AI-AND-LLM.md guide covers prompt injection, tool-argument injection, excessive agency, MCP trust inheritance, and insecure output rendering. Findings must demonstrate a crossed trust boundary, not just model misbehavior.

How does the audit avoid false positives in its report?▼

Every finding passes adversarial validation where a separate agent tries to disprove it, then independent verification where a fresh agent checks each file, line number, and payload against the actual source. Unverifiable claims are rejected or marked as requiring deployment testing.

What output formats does the security audit produce?▼

It writes REPORT.md with an executive summary and findings table, FINDINGS-DETAIL.md with full data flows for medium-and-above findings, and findings.json conforming to a strict schema validated by the bundled validate-findings.cjs script.

Why does one audit run miss some vulnerabilities?▼

Each run explores different code paths depending on what agents discover, and testing shows a single run finds roughly half of total vulnerabilities. The skill reads prior runs' findings.json to skip known issues and target coverage gaps on subsequent runs.

Related Skills

Explore More Software Engineering Skills→
adikpbadikpb
1

clonedeps

Clone dependency source repositories into an ignored local workspace for agent inspection.

Community
Intermediate
jassicsjassics
6

sast-review

Triage SAST findings to rank true vulnerabilities and suggest remediation.

Community
Advanced
HomericIntelligenceHomericIntelligence
18

scan-vulnerabilities

Scan code and dependencies for security vulnerabilities using Bandit, Safety, and Semgrep.

Official
Intermediate
bs-koobs-koo

auditing-web-application-security

Orchestrates static and dynamic security audits of Spring and JSP web applications into unified reports.

Community
Advanced