Pastewatch MCP

Secret redaction MCP server for OpenClaw agents. Prevents API keys, DB credentials, SSH keys, emails, IPs, JWTs, and 30+ other secret types from leaking to L...

This skill appears to do what it claims (local secret redaction and an API proxy), but it requires installing and running a third‑party binary with system-level privileges and intercepting all outbound agent traffic. Before using it: 1) verify the GitHub repository and release artifacts (check signatures/checksums and confirm the release owner), 2) prefer installation via your distro or a vetted package manager if available (brew is suggested for macOS), 3) run the proxy in a restricted environment (dedicated service user, container, or VM) and limit network access, 4) review the pastewatch source code or audits if you can, 5) be aware the setup changes chainwatch/upstream settings and systemd units — plan rollback and backups, and 6) do not install if you cannot fully trust or validate the binary that will process all agent outbound data. If you want a lower‑risk test, run pastewatch locally in a container and exercise its behavior on non-sensitive test data first.

Type: OpenClaw Skill Name: pastewatch-mcp Version: 1.3.0 The pastewatch-mcp skill bundle provides a secret redaction utility that requires high-risk system modifications, including installing binaries to `/usr/local/bin` via curl and creating systemd services to run an API proxy that intercepts outbound LLM traffic. It also includes broad data-access capabilities such as scanning local files, git history, and entire GitHub organizations (via `pastewatch-cli posture`). While these features are consistent with the stated goal of preventing credential leaks, the extensive system control, network interception, and data-gathering potential represent a significant attack surface for an AI agent.