Skill Detail
Block agent egress, MCP prompt injection, and secret exfiltration before agents touch the open internet with Pipelock
Put an inline firewall and containment layer in front of agent network traffic, tool calls, and MCP traffic before you trust an agent with local secrets.
Security & VerificationMulti-Framework
Security & Verification
Multi-Framework
Security Reviewed
⭐ 333 GitHub stars
COPY SKILL INSTRUCTIONS (OPTIONAL)
npx skills add agentskillexchange/skills --skill block-agent-egress-mcp-prompt-injection-and-secret-exfiltration-before-agents-touch-the-open-internet-with-pipelock
Uses the third-party skills CLI, not an ASE-owned installer. Check your agent’s compatibility. This copies instructions; complete the upstream tool setup below separately.
At a glance
Tools required
Homebrew or Go, terminal, supported agent runtime or IDE integration
Install & setup
Install with
brew install luckyPipewrench/tap/pipelock or go install github.com/luckyPipewrench/pipelock/cmd/pipelock@latest, run pipelock init, then place the agent behind pipelock sandbox or pipelock mcp proxy with your policy config.Author
luckyPipewrench
Publisher
Individual
Last updated
Apr 16, 2026
Quick brief
Use Pipelock when the problem is not just running an agent, but safely containing what that agent can send, fetch, and execute once it has credentials or shell access. It sits at the boundary between the agent and the outside world, scans outbound and inbound traffic, wraps MCP servers, and enforces pre-execution policy before risky actions land. The scope boundary is clear: this is an operator workflow for guarding live agent egress and tool use, not a generic security product card or broad SDK listing.