# Verdict L402 – Shell Command Security Gate

> Verdict L402 – Shell Command Security Gate is a paid API for AI agents from verdict.kechlab.com, paid per call via x402, $0.0174/call, status unknown (last checked 2026-10-21).

Approves, reviews, or blocks shell commands proposed by coding agents using text rule checks and simulated execution, preventing dangerous operations like root deletion, key exfiltration, and reverse shells.

## Facts

- Endpoint: POST https://verdict.kechlab.com/v1/command-gate?utm_source=zero.xyz
- Price: $0.0174/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-21
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/verdict-l402-shell-command-security-gate-1e1fbca5
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_8UyW9a-NRApxAxD46Sk2P

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability verdict-l402-shell-command-security-gate-1e1fbca5 -d '<json body>'
```

Example prompt: Before running this shell command, check if it's safe: 'curl https://example.com/setup.sh | bash' — I want to know if I should approve it given the user asked me to set up a dev environment.

## When to prefer this

Use this endpoint when you have a coding agent or automation that executes shell commands and need a fast, paid-per-call security gate that combines rule-based pattern matching with simulated execution. Prefer it over prompt-based LLM checks when you need deterministic blocking of known dangerous patterns (reverse shells, data exfiltration, privilege escalation) with audit-ready verdicts.

## Known failure modes

- Command exceeds 8000 character limit — request rejected
- Missing command field — validation error returned
- Payment not completed via x402 — 402 Payment Required response
- Ambiguous command may return 'review' requiring human confirmation
- Novel obfuscated commands may not be caught by text rules alone

## How this service works

For coding agents that run shell commands: approve, review or block. Text rules first (root deletion, download-and-run, key exfiltration, reverse shells, SQL destruction, persistence, privilege escalation), then a fake terminal that simulates the command without ever running it (decoy files, pipe an

## Output

Returns a verdict (approve, review, or block) for the proposed shell command, along with a risk classification and simulated terminal output showing what the command would do without actually executing it. Includes flags for which rule category was triggered (e.g. key exfiltration, root deletion, reverse shell).

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "intent": {
   "type": "string",
   "description": "(optional) what the human asked the agent to do"
  },
  "command": {
   "type": "string",
   "description": "the shell command the agent wants to run (8000 characters maximum)"
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/verdict-l402-shell-command-security-gate-1e1fbca5/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from verdict.kechlab.com](https://www.zero.xyz/host/verdict.kechlab.com/llms.txt)