Git-Fabric CVE
Provides tools for autonomous CVE detection, enrichment, and remediation across managed repositories using GHSA and NVD data. It enables automated triage and pull request creation for dependency fixes based on configurable severity policies.
README
<p align="center"> <img src="cve-banner.svg" alt="@git-fabric/cve" width="900"> </p>
@git-fabric/cve
CVE detection-to-remediation fabric app. Scan, enrich, triage, and fix vulnerabilities across managed repos — autonomously.
Part of the git-fabric ecosystem.
Architecture
Five composable layers, each independently consumable:
Detection → Intelligence → Decision → Action → State
│ │ │ │ │
│ Scan deps │ NVD enrich │ Policy │ Branch │ JSONL queue
│ query GHSA │ CVSS/CWE │ triage │ commit │ dedup
│ │ │ plans │ PR │ stats
└───────────────┴──────────────┴────────────┴──────────┴──────────
| Layer | What it does | Side effects? |
|---|---|---|
| Detection | Reads dependency manifests, queries GitHub Advisory Database | No (produces findings) |
| Intelligence | Enriches CVEs from NVD with CVSS scores, status, CWE | No (pure data transform) |
| Decision | Applies severity policy, produces triage plans | No (pure logic) |
| Action | Creates branches, commits dependency bumps, opens PRs | Yes (writes to GitHub) |
| State | Manages the CVE queue (JSONL), dedup, filtering, stats | Yes (writes to state repo) |
Quick Start
As an MCP Server
# Set environment
export GITHUB_TOKEN="ghp_..."
export STATE_REPO="ry-ops/git-steer-state"
export MANAGED_REPOS="ry-ops/git-steer,ry-ops/blog"
# Start MCP server (stdio)
npx @git-fabric/cve start
Claude Desktop Config
{
"mcpServers": {
"git-fabric-cve": {
"command": "npx",
"args": ["@git-fabric/cve", "start"],
"env": {
"GITHUB_TOKEN": "ghp_...",
"STATE_REPO": "ry-ops/git-steer-state",
"MANAGED_REPOS": "ry-ops/git-steer,ry-ops/blog"
}
}
}
}
CLI
# Scan repos for vulnerable deps
fabric-cve scan --severity-threshold HIGH
# Enrich a single CVE from NVD
fabric-cve enrich CVE-2024-45519
# Triage pending queue entries (dry run)
fabric-cve triage --dry-run true
# Queue operations
fabric-cve queue list --status pending
fabric-cve queue stats
MCP Tools
| Tool | Description |
|---|---|
cve_scan |
Scan managed repos for vulnerable dependencies via GHSA |
cve_enrich |
Fetch enriched details for a CVE from NVD |
cve_batch |
Batch enrich and rank multiple CVEs by severity |
cve_triage |
Process pending queue entries and open PRs per policy |
cve_queue_list |
List queue entries filtered by status/severity |
cve_queue_stats |
Queue health dashboard |
cve_queue_update |
Manually update entry status (e.g. skip with reason) |
Severity Policy
The decision layer applies configurable policy:
| Severity | Default Action | PR Type |
|---|---|---|
| CRITICAL | Auto-PR | Confirmed |
| HIGH | Auto-PR | Draft |
| MEDIUM | Skip | Manual review |
| LOW | Skip | Noise reduction |
Override via CLI flags or MCP tool arguments:
fabric-cve triage \
--auto-pr-threshold CRITICAL \
--max-prs-per-run 3 \
--require-patched-version true
GitHub Actions
Two workflows for autonomous operation:
cve-scan.yml— Weekly Monday scan, queries GHSA for all managed repos, queues findingscve-triage.yml— Dispatch-only, reads queue, applies policy, opens PRs
The scan explicitly dispatches triage after completing — no push-trigger race condition.
Required Secrets & Variables
| Name | Type | Description |
|---|---|---|
GIT_FABRIC_TOKEN |
Secret | GitHub token with repo + workflow access |
NVD_API_KEY |
Secret | NVD API key (optional, raises rate limit 5→50 req/30s) |
STATE_REPO |
Variable | State repo path (e.g. ry-ops/git-steer-state) |
MANAGED_REPOS |
Variable | Comma-separated repos to manage |
Consuming from git-steer
git-steer can delegate to the fabric by implementing the GitHubAdapter and StateAdapter interfaces:
import { layers } from "@git-fabric/cve";
// Detection
const result = await layers.detection.detect(repos, "HIGH", githubAdapter);
// Queue
await layers.state.enqueue(result.findings, stateAdapter);
// Triage
const pending = await layers.state.pending(stateAdapter);
const plans = layers.decision.triage(pending, policy);
const results = await layers.action.execute(plans, githubAdapter);
Project Structure
src/
├── types.ts # Shared types + adapter interfaces
├── index.ts # Barrel export
├── layers/
│ ├── detection.ts # GHSA scanning + manifest parsing
│ ├── intelligence.ts # NVD enrichment
│ ├── decision.ts # Severity policy engine
│ ├── action.ts # Branch + commit + PR creation
│ └── state.ts # JSONL queue management
├── mcp/
│ └── server.ts # MCP server (7 tools)
└── adapters/
└── env.ts # Env var → Octokit adapter
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。