Articles
20 articles. Each has copy-paste commands and cited sources. Each is also Markdown at the same path plus .md, and inside llms-full.txt.
Runtime guides
One per agent runtime.
- Secure an OpenClaw agentRegister OpenClaw with SMA, run the Checkup, lock the gateway, and add the heartbeat.
- Secure a Claude Code agentRegister Claude Code with SMA, tighten permissions, vet MCP servers, and schedule the heartbeat.
- Secure a Hermes agentRegister a Hermes agent with SMA, run the Checkup from its terminal tool, and schedule the heartbeat.
- Secure a Codex agentRegister Codex CLI with SMA, set the sandbox and approval mode, vet MCP servers in config.toml, and add the heartbeat.
- Secure a Cursor agentRegister Cursor with SMA, audit mcp.json, turn off blind auto-run, and add the heartbeat.
- Secure a custom MCP clientOnboard any MCP client with SMA over HTTP, and the client-side rules that stop tool poisoning.
- Secure an n8n agent workflowRegister an n8n workflow as an SMA agent, lock down the n8n host, and run the heartbeat from a Schedule trigger.
Fix guides
One per Checkup check.
- Fix: your gateway is on 0.0.0.0Find agent services listening on all interfaces and bind them to localhost.
- Fix: your runtime or MCP server has a known CVEFind vulnerable versions of your agent runtime and MCP packages, and update them.
- Fix: plaintext API keys in config filesFind keys in configs, .env files and memory files. Move them out. Rotate them.
- Fix: a skill or MCP server has a bad verdictRemove skills and MCP servers that are malicious or unknown, and vet new ones before install.
- Fix: shell, network and file-write with no approvalCut tool permissions to what the agent needs and require approval for risky calls.
- Fix: no egress allowlistLimit which hosts the agent can reach so a hijacked agent cannot send your data out.
- Fix: no stable identity and no audit logGive the agent a stable DID and log every tool call.
- Fix: untrusted content in long-term memoryStop web and tool content from becoming trusted memory. Record where each memory came from.
Threat explainers
What happened, and what to run.
- ClawHavoc - how malicious skills reached OpenClaw agents341 of 2,857 audited ClawHub skills were malicious. What they did and how to check yours.
- Exposed gateways and CVE-2026-25253OpenClaw gateways on port 18789 bound to 0.0.0.0 with auth off. A one-click RCE with CVSS 8.8.
- MCP tool poisoning and rug pullsHidden instructions in tool descriptions, and servers that change after you approved them.
- The Moltbook token leakA Supabase key in client JavaScript with row-level security off exposed about 1.5M agent tokens.
- Prompt-injection exfiltrationHow a web page, form or ticket makes an agent send your data to an attacker, and how to cut the path.