Shrike Security
Run AI-agent-focused security checks—prompt injection, SQL injection, PII handling, and threat intel—from your coding agent before you ship.
Overview
Shrike Security is an MCP server for the Ship phase that scans AI agent integrations for prompt injection, SQL injection, PII isolation issues, and related threat intel.
What is this MCP server?
- Prompt injection detection tailored to AI agent surfaces
- SQL injection checks for data paths your agent touches
- PII isolation analysis to reduce accidental exposure in tool outputs
- Threat intelligence signals packaged for MCP via npx shrike-mcp (v1.1.1 stdio)
- npm package shrike-mcp version 1.1.1
- Documented capability areas: prompt injection, SQL injection, PII isolation, threat intel
- stdio transport with npx runtime hint
Community signal: 1 GitHub stars.
What problem does it solve?
Agent features introduce new attack surfaces, and solo builders lack a fast way to security-scan prompts, tools, and queries from inside the IDE.
Who is it for?
Solo developers shipping LLM agents, MCP tools, or RAG APIs who want automated security signals in the same session as code review.
Skip if: Non-AI CRUD apps with no model or tool calling, or teams that require a full SOC2-grade external pentest only.
What do I get? / Deliverables
After you add Shrike, your agent can surface injection, leakage, and threat findings during pre-ship security passes on agent code.
- Security scan results for agent-facing code paths
- Documented findings on injection, PII, and threat indicators
- Pre-ship confidence pass integrated into agent-driven review
Recommended MCP Servers
Journey fit
How it compares
Agent-focused security MCP scanner, not a general dependency CVE dashboard or hosting firewall.
Common Questions / FAQ
Who is Shrike Security for?
Builders shipping AI agents and MCP integrations who need prompt injection, SQLi, and PII checks callable from their coding agent.
When should I use Shrike Security?
Use it in Ship before release or after major agent/tool changes when you are reviewing security on branches and PRs.
How do I add Shrike Security to my agent?
Configure stdio MCP with npx and package shrike-mcp per server metadata, then run scans from your MCP-enabled client during review.