菜单
AI-powered code vulnerability analysis and dependency scanning using Gemini CLI security extension patterns. Detects hardcoded secrets, injection attacks, weak cryptography, authentication flaws, and LLM prompt injection. Also scans dependencies against the OSV.dev vulnerability database.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/oimiragieo/agent-studio --skill gemini-cli-security命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Agent frontmatter enhancements — disallowedTools, mcpServers scoping, fork_eligible field
Context management — pre-compact persistence, threshold alignment, microcompact/circuit breaker detection
Hook enhancements — updatedInput for bash safety prefixes, suppressOutput for verbose blocks, denial-based routing feedback
正在显示 SKILL.md
基于 SOC 职业分类
| name | gemini-cli-security |
| description | AI-powered code vulnerability analysis and dependency scanning using Gemini CLI security extension patterns. Detects hardcoded secrets, injection attacks, weak cryptography, authentication flaws, and LLM prompt injection. Also scans dependencies against the OSV.dev vulnerability database. |
| version | 1.0.0 |
| model | sonnet |
| invoked_by | both |
| user_invocable | true |
| tools | ["Read","Write","Bash","WebFetch","Glob","Grep"] |
| args | [target-path] [--scan-deps] [--json] [--scope <natural-language-scope>] |
| best_practices | ["Analyze code changes in pull requests to catch vulnerabilities early","Always check dependencies for CVEs via OSV.dev integration","Produce structured JSON output for automated pipelines","Flag LLM-specific risks (prompt injection, unsafe output handling)","Report precision and recall benchmarks when assessing coverage"] |
| error_handling | graceful |
| streaming | supported |
| verified | false |
| lastVerifiedAt | "2026-02-19T05:29:09.098Z" |
| source | builtin |
| trust_score | 100 |
| provenance_sha | bef705454e10ccb7 |
This skill adapts the Gemini CLI Security Extension's analysis methodology for the agent-studio framework. The original extension uses two MCP server patterns — a security analysis server and an OSV-Scanner integration — to provide dual-vector coverage. This skill implements equivalent analysis using native Claude Code tools (WebFetch for OSV.dev API, Grep/Bash for static analysis patterns).
Source repository: https://github.com/gemini-cli-extensions/security
License: Apache 2.0
Performance: 90% precision, 93% recall (OpenSSF CVE benchmark, TypeScript/JavaScript)
NO PRODUCTION CODE WITHOUT SECURITY ANALYSIS FOR AUTH/SECRETS/EXTERNAL-INPUT HANDLERS
All code paths handling authentication, hardcoded values, external input, or AI model outputs MUST be analyzed before production deployment.
| Pattern | Detection Method |
|---|---|
| Hardcoded API keys | Grep for key patterns + entropy analysis |
| Hardcoded passwords | Credential keyword detection |
| Private keys in source | PEM block / base64 key detection |
| Encryption keys | Symmetric key constant patterns |
| Attack Type | Examples |
|---|
| SQL injection | String concatenation in queries |
| XSS | Unescaped user content in HTML/JS output |
| Command injection | Shell exec with user-controlled args |
| SSRF | User-controlled URLs in server requests |
| Template injection | Unsanitized user input in template engines |
| Flaw | Detection |
|---|---|
| Session bypass | Missing auth middleware |
| Weak tokens | Predictable token generation |
| Insecure password reset | Token-less or email-only resets |
| Missing MFA enforcement | Auth flows without 2FA checks |
| Issue | Detection |
|---|---|
| Weak cryptography | MD5/SHA1 for secrets; DES/RC4 usage |
| Sensitive data in logs | PII/credential patterns in log statements |
| PII violations | Unencrypted PII storage or transmission |
| Insecure deserialization | Unsafe pickle/eval/deserialize calls |
| Risk | Detection |
|---|---|
| Prompt injection | User content injected into LLM prompts without sanitization |
| Unsafe output handling | LLM output used in exec/eval/shell without validation |
| Insecure tool integration | Tool calls with unchecked LLM-provided parameters |
// From an agent
Skill({ skill: 'gemini-cli-security' });
// With arguments via Bash integration
Skill({ skill: 'gemini-cli-security', args: 'src/ --scan-deps' });
# Analyze code in a directory
node .claude/skills/gemini-cli-security/scripts/main.cjs --target src/
# Scan dependencies for CVEs
node .claude/skills/gemini-cli-security/scripts/main.cjs --scan-deps
# JSON output for CI integration
node .claude/skills/gemini-cli-security/scripts/main.cjs --target . --json
# Scoped analysis with natural language
node .claude/skills/gemini-cli-security/scripts/main.cjs --target src/auth/ --scope "focus on token handling and session management"
Default output (markdown report):
## Security Analysis Report
### CRITICAL
- [AUTH-001] Hardcoded API key found in src/config.ts:42
Pattern: `const API_KEY = "sk-..."`
Remediation: Move to environment variable
### HIGH
- [INJ-002] SQL injection risk in src/db/users.ts:87
Pattern: String concatenation in query builder
Remediation: Use parameterized queries
### Dependencies
- lodash@4.17.15 → CVE-2021-23337 (HIGH) - Prototype pollution
Fix: Upgrade to lodash@4.17.21+
JSON output (--json flag):
{
"findings": [
{
"id": "AUTH-001",
"severity": "CRITICAL",
"category": "secrets",
"file": "src/config.ts",
"line": 42,
"description": "Hardcoded API key",
"remediation": "Move to environment variable"
}
],
"dependencies": [
{
"package": "lodash",
"version": "4.17.15",
"cve": "CVE-2021-23337",
"severity": "HIGH",
"fix": "4.17.21"
}
],
"summary": {
"critical": 1,
"high": 2,
"medium": 3,
"low": 0,
"precision": 0.9,
"recall": 0.93
}
}
The skill integrates with the OSV.dev API (no authentication required) to check dependencies:
// OSV.dev batch query endpoint
WebFetch({
url: 'https://api.osv.dev/v1/querybatch',
prompt: 'Extract vulnerability IDs, severity, and affected versions for these packages',
});
Supported ecosystems: npm, PyPI, RubyGems, Maven, Go, Cargo, NuGet, Packagist
The original extension supports PR analysis via GitHub Actions. This skill includes an equivalent workflow template:
# .github/workflows/security.yml
name: Security Analysis
on: [pull_request]
jobs:
security:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- name: Run security analysis
run: node .claude/skills/gemini-cli-security/scripts/main.cjs --target . --json
Why native tools over MCP servers: The original extension uses two MCP servers (security analysis server + OSV-Scanner binary). This skill uses native Claude Code tools instead:
Deviation from source: The original uses Gemini AI for code analysis; this skill uses the pattern-based detection methodology documented in the extension's benchmarking. The AI analysis component can be provided by the invoking agent (security-architect) rather than an embedded AI call.
| Agent | Role |
|---|---|
security-architect | Primary: comprehensive security audits |
developer | Supporting: pre-commit security checks |
code-reviewer | Supporting: PR review security layer |
Before starting:
Read .claude/context/memory/learnings.md
After completing:
.claude/context/memory/learnings.md.claude/context/memory/issues.md.claude/context/memory/decisions.mdASSUME INTERRUPTION: If it's not in memory, it didn't happen.