搜索「Prompt Injection」的结果

LLM 可观测性与编排智能体

io.smithery.a2a.llm-observability-orchestration

通过 LangChain（system + human）链在 Vertex AI 的 Gemini 上运行提示词，可选开启 LangSmith 追踪。适合需要追踪与可观测的 LangChain 流水线、提示词评估与链质量监控。

Compuute MCP Security Scanner

io.smithery.daniel-abbay.compuute-scan-api

Static security scanner for MCP servers. POST a public GitHub URL, get severity counts, a score, and the top findings with file+line back. 37 rules across TypeScript, JavaScript, Python, Go, Rust, C#, Java, and Kotlin — every language with an official MCP SDK. Detects argument injection for npx/uvx/pipx/pnpx runner binaries (CWE-88), known CVEs in 40+ top packages, and the usual L0 discovery (transport, tool inventory, dependency pinning). This is a pattern detector, not an exploitability oracle. Around 90% raw false-positive rate on unfiltered output — triage is on you, and the response says so explicitly. POST /v1/scan is free with no API key. POST /v1/scan/pay charges $0.10 USDC per scan via x402 on Base. Manual L2-L4 audits at compuute.se/audit when you need dataflow review. Wraps compuute-scan (MIT, zero deps). Per-rule false-positive rates and the methodology paper live in the repo.

io.github.getmcpm/cli

MCP 安全守卫兼包管理器：按信任评分安装，拦截提示注入与跑路（rug-pull）。

com.runningunattended/agent-lint

Checks CLAUDE.md, AGENTS.md or a system prompt for 12 safeguards unattended agents need.

io.github.letdraw/letdraw

Create, edit, share and export hand-drawn LetDraw diagrams, from code or a prompt.

ai.mails/agent-email

A real inbox for AI agents: send, receive and thread email, behind a prompt-injection firewall.

io.github.llschmidt/meerkat

A prompt engineer you talk to. It writes prompts, sharpens yours, and keeps them on your shelf.

io.github.tylerscomic-lab/github-actions-audit-mcp

Audit GitHub Actions workflows for script injection, unpinned actions and missing permissions.

io.github.tylerscomic-lab/agent-skill-audit-mcp

Scan AI agent skills and configs for hidden Unicode, prompt injection and exfiltration.

ai.blipsmith/sound-effects

Game sound effects from a text prompt: game-ready WAVs with seamless loops and variations.

com.ismalicious/mcp-server

面向 AI 智能体的指标信誉判定、CVE 查询（CVSS、EPSS、KEV）与提示注入扫描。

io.github.abyssbugg-labs/sourcery

Sourcery security findings: MCP server, CLI, and fix-prompt builder over the Sourcery API.

io.github.PlusDocs/plus-ai-mcp

Create professionally designed, editable PowerPoint presentations from a prompt.

io.github.abyssbugg/sourcery

Sourcery security findings: MCP server, CLI, and fix-prompt builder over the Sourcery API.

io.github.4hmetuyar/mcp-server-auditor

Scans MCP server tool definitions for excessive agency, injection sinks, hardcoded secrets

io.github.4hmetuyar/ai-code-scanner

Scans code for insecure LLM/AI integration: exposed keys, unsafe output, prompt injection

io.github.loopg/pramana-mcp

Read your recorded AI agent runs, and what changed when you swapped a model or prompt.

io.github.oidsdev/oids

MCP endpoint for Oids, a microblogging network for AI agents: posts, DMs, prompt packs.

io.github.ashlrai/lexicon

Fixes the names and jargon speech-to-text gets wrong before your agent acts on a dictated prompt.

com.loomaly/loomaly

SEO audit of every page: what to fix first, a fix prompt for your framework, and a re-check.

io.github.nzdsf2-gif/relayshield-mcp

数据泄露、SIM 卡调换、信息窃取器、域名仿冒、MCP 注册表风险与提示注入检测。

io.github.gamaze-labs/hicortex

Shared fleet memory for AI agents: nightly self-correction, recall every prompt (supported agents).

co.pixelcove/citation-tracker

检查 AI 搜索引擎在回答某个提示时是否引用你的目标域名，通过 MCP 调用 $0.75/次。

com.bunstation/bunstation

Build websites and apps from a prompt, edit files, publish, and check credits and usage.