agent-security-scanner

Agentic security patterns for AI agent systems including attack vector defense, sandboxing, input sanitization, security scanning, CVE awareness, and least-privilege tool access. Use when.

Ir para a instalação

Repositório: oyi77/1ai-skills
Última atividade na origem: 31 de julho de 2026 às 15:22
Idioma detectado do SKILL.md: inglês
Estrelas: 8
Forks: 0

Opções de instalação

Por padrão, está selecionado o prompt que primeiro revisa a origem. Você pode mudar para um comando direto ou baixar uma cópia local.

Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.

Um comando direto ignora o prompt de revisão. Verifique a origem antes de executá-lo.

npx skills add https://github.com/oyi77/1ai-skills --skill agent-security-scanner

O comando permanece em uma só linha. Role horizontalmente para revisá-lo antes de copiar.

Prefere uma cópia local? Baixe os arquivos disponíveis atualmente no SkillsMP.

Revise os arquivos de origem

Leia o SKILL.md e os arquivos complementares exibidos pelo SkillsMP antes de decidir se vai instalar.

SKILL.md

Instruções da origem · Visualização somente leitura

Voltar à pesquisa Voltar à instalação

mobile-hacking

Android and mobile application security testing — emulators, rooting, traffic interception, dynamic instrumentation. Use when testing mobile apps for vulnerabilities, reversing APKs, or bypassing security controls on Android.

2026-08-10

self-improving

Self-reflection + Self-criticism + Auto-learning from corrections + Self-organizing memory. Agent evaluates its own work, catches mistakes, and improves permanently. Use when working with self improving.

2026-08-09

conducting-full-scope-red-team-engagement

Plan and execute a comprehensive red team engagement covering reconnaissance through post-exploitation using MITRE ATT&CK-aligned TTPs to evaluate an organization's detection and response capabilities. Use when working with conducting full scope red team engagement.

2026-08-09

Ver todas deste repositório

Rationalization

Reality

"I'll figure it out as I go"

A structured approach saves time and reduces errors. Follow the workflow in this skill rather than improvising.

"I already know this topic"

Familiarity breeds shortcuts. Use the checklist to verify you haven't missed critical steps.

"This doesn't apply to my situation"

The patterns here generalize across contexts. Adapt, don't skip — the underlying principles hold.

"One more tool will fix it"

Adding complexity rarely solves process gaps. Master the core workflow first.

Rationalization

Reality

"I'll figure it out as I go"

A structured approach saves time and reduces errors. Follow the workflow in this skill rather than improvising.

"I already know this topic"

Familiarity breeds shortcuts. Use the checklist to verify you haven't missed critical steps.

"This doesn't apply to my situation"

The patterns here generalize across contexts. Adapt, don't skip — the underlying principles hold.

"One more tool will fix it"

Adding complexity rarely solves process gaps. Master the core workflow first.

name	agent-security-scanner
description	Agentic security patterns for AI agent systems including attack vector defense, sandboxing, input sanitization, security scanning, CVE awareness, and least-privilege tool access. Use when.
domain	core
author	oyi77
license	Apache-2.0
subdomain	core-platform
tags	["security","prompt-injection","sandboxing","agentshield","least-privilege","cve","scanning"]
persona	{"name":"AgentShield","title":"Agentic Security Engineer","expertise":["AI Agent Attack Vectors","Sandboxing","Prompt Injection Defense","Tool Permission Hardening","CVE Analysis","Security Scanning"],"philosophy":"Trust nothing. Verify everything. Sandbox what you cannot trust. Limit what you cannot sandbox."}
version	1.0.0

name	agent-security-scanner
description	Agentic security patterns for AI agent systems including attack vector defense, sandboxing, input sanitization, security scanning, CVE awareness, and least-privilege tool access. Use when.
domain	core
author	oyi77
license	Apache-2.0
subdomain	core-platform
tags	["security","prompt-injection","sandboxing","agentshield","least-privilege","cve","scanning"]
persona	{"name":"AgentShield","title":"Agentic Security Engineer","expertise":["AI Agent Attack Vectors","Sandboxing","Prompt Injection Defense","Tool Permission Hardening","CVE Analysis","Security Scanning"],"philosophy":"Trust nothing. Verify everything. Sandbox what you cannot trust. Limit what you cannot sandbox."}
version	1.0.0

Vector	Attack	Defense
CLAUDE.md / AGENTS.md	Hidden instructions in project files override agent behavior	Scan for auto-run directives, hidden text, conflicting instructions
User input	Crafted prompts bypass safety guardrails	Input sanitization, instruction hierarchy enforcement
File content	Malicious code comments or strings inject instructions	Content scanning before agent reads untrusted files
MCP responses	MCP server returns injected instructions	Validate MCP response format, strip instruction-like content
Subagent output	Subagent returns instructions instead of data	Treat subagent output as data, never as instructions

Vector	Attack	Defense
Bash execution	Agent runs arbitrary shell commands from untrusted input	Sandbox with allowlists, never interpolate untrusted input into commands
File write	Agent writes malicious files (backdoors, exfiltration scripts)	Restrict write paths, validate file content before write
File read	Agent reads sensitive files (credentials, secrets, keys)	Denylist for sensitive paths (.env, credentials, private keys)
Network access	Agent makes outbound requests to exfiltrate data	Block outbound HTTP from hooks, validate URLs in tool calls
MCP tools	Agent calls MCP tools with malicious parameters	Validate MCP tool parameters, rate-limit tool calls

Vector	Attack	Defense
Commit content	Secrets committed to git	Pre-commit secret scanning (gitleaks, truffleHog)
Tool output	Agent includes sensitive data in tool responses	Redact sensitive patterns from tool outputs
Subagent delegation	Sensitive context passed to subagent that logs it	Minimize context, exclude secrets from subagent payloads
Hook logs	Hooks log sensitive data to files	Sanitize hook output, exclude secrets from logs

agent-security-scanner

Opções de instalação

Revise os arquivos de origem

Overview

Anti-Rationalization Table

When to Use

When NOT to Use

Overview

Anti-Rationalization Table

When to Use

When NOT to Use

Process

1. Attack Vector Inventory

2. Sandboxing Untrusted Code Execution

3. Input Sanitization for Agent Inputs

4. Security Scanning Integration

5. CVE Awareness for Agent Dependencies

6. Principle of Least Privilege for Agent Tools

Common Rationalizations

Red Flags

Verification

Component	Checks
CLAUDE.md / AGENTS.md	Hardcoded secrets, auto-run instructions, prompt injection patterns
settings.json / config.json	Overly permissive allow lists, missing deny lists, dangerous bypass flags
mcp.json	Risky MCP servers, hardcoded env secrets, npx supply chain risks
hooks/	Command injection via interpolation, data exfiltration, silent error suppression
agents/*.md	Unrestricted tool access, prompt injection surface, missing model specs
Package dependencies	Known CVEs, outdated packages, suspicious dependencies

Tier	Tools Allowed	Use Case
Read-Only	Read, Grep, Glob, LSP tools	Code exploration, research
Standard	Read-Only + Edit, Write, Bash (allowlisted)	Daily development
Admin	Standard + unrestricted Bash, file system	Infrastructure, CI/CD
Superuser	All tools including MCP admin tools	Emergency only, requires approval

Rationalization	Reality
"My agent only runs locally, security is not a concern"	Local agents still read untrusted files, execute code, and can exfiltrate data via git commits or network calls
"Prompt injection is theoretical, not a real threat"	Real-world prompt injection attacks have been documented in production agent systems; OWASP ranks it as a top LLM risk
"The agent is smart enough to avoid dangerous commands"	Agents follow instructions, including injected ones; defense must be at the harness layer, not the model layer
"I trust all my MCP servers"	Supply chain attacks compromise trusted packages; trust but verify with pinned versions and dependency audits
"Scanning slows down my workflow"	A 10-second scan prevents hours of incident response; automate it in CI to keep the workflow fast
"Allowlists are too restrictive"	Start tight, loosen only when a specific legitimate use is blocked; the inverse (start loose, try to tighten) never works

agent-security-scanner

Informações da origem

Opções de instalação

Revise os arquivos de origem

Mais deste repositório

Mais deste repositório

Overview

Anti-Rationalization Table

When to Use

When NOT to Use

Overview

Anti-Rationalization Table

When to Use

When NOT to Use

Process

1. Attack Vector Inventory

2. Sandboxing Untrusted Code Execution

3. Input Sanitization for Agent Inputs

4. Security Scanning Integration

5. CVE Awareness for Agent Dependencies

6. Principle of Least Privilege for Agent Tools

Common Rationalizations

Red Flags

Verification