{
  "markdown": "[![CI](https://github.com/acebot712/promptguard-plugin/actions/workflows/ci.yml/badge.svg)](https://github.com/acebot712/promptguard-plugin/actions/workflows/ci.yml)\n[![License](https://img.shields.io/github/license/acebot712/promptguard-plugin)](https://github.com/acebot712/promptguard-plugin/blob/main/LICENSE)\n\n# PromptGuard Agent Skills\n\nLLM security for AI coding agents — protect applications from prompt injection, PII leakage, and data exfiltration. Works with **Cursor, Claude Code, Codex, Copilot, Gemini CLI, Windsurf, VSCode**, and any MCP-compatible agent.\n\n## Installing the MCP Server\n\nPromptGuard ships a native MCP server built into the CLI. Install the CLI first, then add the MCP server to your agent.\n\n### Prerequisites\n\n- **PromptGuard CLI**:\n  - macOS: `brew install promptguard/tap/promptguard`\n  - Linux/macOS: `curl -fsSL https://raw.githubusercontent.com/acebot712/promptguard-cli/main/install.sh | sh`\n  - Cargo: `cargo install promptguard`\n- **PromptGuard account** — sign up at [promptguard.co](https://promptguard.co)\n- **API key** — get one from [app.promptguard.co/settings/api-keys](https://app.promptguard.co/settings/api-keys)\n\n### Per-agent setup\n\n<details>\n<summary>Install in <b>Cursor</b></summary>\n<br />\n\n**One-click install** (opens Cursor and installs the plugin + MCP server):\n\n[Install PromptGuard in Cursor](cursor://anysphere.cursor-deeplink/mcp/install?name=promptguard&config=eyJjb21tYW5kIjoicHJvbXB0Z3VhcmQiLCJhcmdzIjpbIm1jcCIsIi10Iiwic3RkaW8iXX0=)\n\nOr add to `.cursor/mcp.json`:\n\n```json\n{\n  \"mcpServers\": {\n    \"promptguard\": {\n      \"command\": \"promptguard\",\n      \"args\": [\"mcp\", \"-t\", \"stdio\"]\n    }\n  }\n}\n```\n\nCursor also supports the plugin's rules, commands, and agent — install via Settings > Plugins > search \"promptguard\".\n\n</details>\n\n<details>\n<summary>Install in <b>Claude Code</b></summary>\n<br />\n\n**One-line MCP install** (fastest — registers the MCP server via the Claude CLI):\n\n```bash\nclaude mcp add promptguard -- promptguard mcp -t stdio\n```\n\n**As a plugin** (includes skills, agents, and the MCP server):\n\n```bash\nclaude --plugin-dir /path/to/promptguard-plugin\n```\n\nTo submit to the official Claude Code marketplace: [claude.ai/settings/plugins/submit](https://claude.ai/settings/plugins/submit)\n\n</details>\n\n<details>\n<summary>Install in <b>VSCode / Copilot</b></summary>\n<br />\n\nAdd to your `.vscode/mcp.json`:\n\n```json\n{\n  \"servers\": {\n    \"promptguard\": {\n      \"command\": \"promptguard\",\n      \"args\": [\"mcp\", \"-t\", \"stdio\"]\n    }\n  }\n}\n```\n\n</details>\n\n<details>\n<summary>Install in <b>Windsurf</b></summary>\n<br />\n\nAdd to `~/.codeium/windsurf/mcp_config.json`:\n\n```json\n{\n  \"mcpServers\": {\n    \"promptguard\": {\n      \"command\": \"promptguard\",\n      \"args\": [\"mcp\", \"-t\", \"stdio\"]\n    }\n  }\n}\n```\n\n</details>\n\n<details>\n<summary>Install in <b>Codex</b></summary>\n<br />\n\n**As a plugin** (includes skills and MCP server):\n\nCopy to your repo plugins directory and register in your marketplace:\n\n```bash\ncp -R /path/to/promptguard-plugin ./plugins/promptguard\n```\n\nAdd to `.agents/plugins/marketplace.json`:\n\n```json\n{\n  \"name\": \"local-plugins\",\n  \"plugins\": [{\n    \"name\": \"promptguard\",\n    \"source\": { \"source\": \"local\", \"path\": \"./plugins/promptguard\" },\n    \"policy\": { \"installation\": \"AVAILABLE\", \"authentication\": \"ON_INSTALL\" },\n    \"category\": \"Security\"\n  }]\n}\n```\n\nOr add the MCP server standalone in `.codex/mcp.json`:\n\n```json\n{\n  \"mcpServers\": {\n    \"promptguard\": {\n      \"command\": \"promptguard\",\n      \"args\": [\"mcp\", \"-t\", \"stdio\"]\n    }\n  }\n}\n```\n\n</details>\n\n<details>\n<summary>Install in <b>Gemini CLI</b></summary>\n<br />\n\nAdd to `~/.gemini/settings.json`:\n\n```json\n{\n  \"mcpServers\": {\n    \"promptguard\": {\n      \"command\": \"promptguard\",\n      \"args\": [\"mcp\", \"-t\", \"stdio\"]\n    }\n  }\n}\n```\n\n</details>\n\n<details>\n<summary>Install in <b>any other MCP client</b></summary>\n<br />\n\nRun the MCP server via stdio:\n\n```bash\npromptguard mcp -t stdio\n```\n\nConfigure your MCP client to spawn `promptguard mcp -t stdio` as a subprocess.\n\n</details>\n\n## MCP Tools\n\nWhen connected, the agent can call these tools directly:\n\n| Tool | Description |\n|------|-------------|\n| `promptguard_auth` | Authenticate with PromptGuard. Opens the dashboard to copy an API key. |\n| `promptguard_logout` | Remove locally stored credentials. |\n| `promptguard_scan_text` | Scan text for prompt injection, jailbreaks, PII leakage, and toxicity. |\n| `promptguard_scan_project` | Scan a project directory for unprotected LLM SDK usage. |\n| `promptguard_redact` | Redact PII (emails, phones, SSNs, credit cards, API keys) from text. |\n| `promptguard_status` | Show current configuration, active providers, and key type. |\n\n## What's included\n\nBeyond the MCP server, this repo includes agent skills and workflows:\n\n| Component | Description |\n|-----------|-------------|\n| **Skills** | **Secure LLM Integration** — walks through adding PromptGuard to any project. **PromptGuard API** — teaches the agent exact request/response schemas for scan, redact, and guard endpoints. |\n| **Commands** | `/promptguard-scan` finds unprotected LLM calls and hardcoded secrets. `/promptguard-secure` adds PromptGuard to a project end-to-end. |\n| **Agent** | LLM Security Reviewer — specialized code reviewer for prompt injection, PII leakage, agent tool abuse, and other LLM-specific vulnerabilities. |\n| **Rules** | Always-on guidance ensuring the agent writes secure LLM code when using OpenAI, Anthropic, or any supported provider. |\n\nThese components work natively in Cursor (via the plugin system). For other agents, the skills and agent definitions use standard `SKILL.md` format compatible with Claude Code, Codex, and any agent that reads markdown skill files.\n\n## Project attribution\n\nTo associate MCP requests with a specific PromptGuard project (for per-project billing and analytics), set the `project_id` in your local `.promptguard.json`:\n\n```json\n{\n  \"version\": \"1.0\",\n  \"api_key\": \"pg_live_xxx\",\n  \"project_id\": \"your-project-id\",\n  \"proxy_url\": \"https://api.promptguard.co/api/v1\"\n}\n```\n\nOr select a project globally:\n\n```bash\npromptguard projects list\npromptguard projects select <project-id>\n```\n\n## Supported LLM providers\n\nOpenAI, Anthropic, Google Generative AI (Gemini), Cohere, AWS Bedrock, LangChain, CrewAI, LlamaIndex, Vercel AI SDK.\n\n## Links\n\n- [Documentation](https://docs.promptguard.co)\n- [Dashboard](https://app.promptguard.co)\n- [CLI](https://github.com/acebot712/promptguard-cli)\n- [Python SDK](https://github.com/acebot712/promptguard-python)\n- [Node.js SDK](https://github.com/acebot712/promptguard-node)\n- [Website](https://promptguard.co)\n",
  "bytes": 6633,
  "sha": "b13e27ed81cf5c781315bce97207ba68c00062f8322ed8d4fbe992b04ee51ce2",
  "repo_slug": "acebot712/promptguard-plugin",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_acebot712_promptguard_plugin_promptguard_700917fe/readme"
}