Vigil

AI agent safety guardrails for tool calls. Use when (1) you want to validate agent tool calls before execution, (2) building agents that run shell commands, file operations, or API calls, (3) adding a safety layer to any MCP server or agent framework, (4) auditing what your agents are doing. Catches destructive commands, SSRF, SQL injection, path traversal, data exfiltration, prompt injection, and credential leaks. Zero dependencies, under 2ms.

Security Scan

The skill's claims (an agent safety wrapper) align with what it includes and requests, but it depends on an external npm package so you should review that package's source before installing.

Purpose & Capability

Name/description match the actual contents: a small CLI wrapper and instructions that call the npm package vigil-agent-safety to evaluate agent tool calls. The dependency on an npm safety library is coherent with the stated purpose.

Instruction Scope

SKILL.md instructs the user to install and use the vigil-agent-safety npm package and shows example usage. The included script only parses args and forwards them to checkAction; it does not read unrelated system files or environment variables. Note: the skill requires the external package to be installed for runtime behavior—if missing the wrapper will error out.

Install Mechanism

No automated install spec is provided; SKILL.md asks you to run npm install for a public npm package. Using a public npm package is a moderate supply-chain risk (expected here), so review the package and its GitHub repo before installing. The install is not via an opaque URL or archive.

Credentials

The skill requests no environment variables, credentials, or config paths. The wrapper and documentation do not reference secrets or unrelated credentials.

Persistence & Privilege

always is false and the skill is user-invocable. The skill does not request permanent presence or modify other skills' configs. It does not attempt to persist credentials or change system-wide settings.

Assessment

This skill is internally consistent: it provides a small wrapper that calls an npm safety library to evaluate agent tool calls. Before installing or running it, review the vigil-agent-safety package and its GitHub repository (maintainers, recent commits, published code) because the runtime behavior depends entirely on that third‑party package. Pin a specific version, inspect its source code, and install in a sandboxed environment if possible. Note the registry metadata at the top of the skill listing lacked a source/homepage while SKILL.md and _meta.json reference hexitlabs — verify that the npm package and repository URLs match and are legitimate. If you cannot audit the package, avoid installing it on sensitive systems.

Vigil — Agent Safety Guardrails

Validates what AI agents DO, not what they SAY. Drop-in safety layer for any tool-calling agent.

Prerequisites

This skill requires the vigil-agent-safety npm package (12.3KB, Apache 2.0 license):

npm install vigil-agent-safety

• Source code: https://github.com/hexitlabs/vigil
• npm: https://www.npmjs.com/package/vigil-agent-safety
• The npm package has zero runtime dependencies. This skill is a wrapper that calls that package.

Quick Start

import { checkAction } from 'vigil-agent-safety';

const result = checkAction({
  agent: 'my-agent',
  tool: 'exec',
  params: { command: 'rm -rf /' },
});

// result.decision === "BLOCK"
// result.reason === "Destructive command pattern"
// result.latencyMs === 0.3

What It Catches

• Destructive commands (rm -rf, mkfs, reverse shells) → BLOCK
• SSRF (metadata endpoints, localhost, internal IPs) → BLOCK
• Data exfiltration (curl to external, .ssh/id_rsa access) → BLOCK
• SQL injection (DROP TABLE, UNION SELECT) → BLOCK
• Path traversal (../../../etc/shadow) → BLOCK
• Prompt injection (ignore instructions, [INST] tags) → BLOCK
• Encoding attacks (base64 decode, eval(atob())) → BLOCK
• Credential leaks (API keys, AWS keys, tokens) → ESCALATE

22 rules. Zero dependencies. Under 2ms per check.

Modes

import { configure } from 'vigil-agent-safety';

// warn = log violations but don't block (recommended to start)
configure({ mode: 'warn' });

// enforce = block dangerous calls
configure({ mode: 'enforce' });

// log = silent logging only
configure({ mode: 'log' });

Use with Clawdbot

Add Vigil as a safety layer for your agent tool calls. The scripts/vigil-check.js wrapper lets you validate from the command line:

# Check a tool call
node scripts/vigil-check.js exec '{"command":"rm -rf /"}'
# → BLOCK: Destructive command pattern

# Check a safe call
node scripts/vigil-check.js read '{"path":"./README.md"}'
# → ALLOW

Policies

Load built-in policy templates:

import { loadPolicy } from 'vigil-agent-safety';

loadPolicy('restrictive');  // Tightest rules
loadPolicy('moderate');     // Balanced (default)
loadPolicy('permissive');   // Minimal blocking

CLI

npx vigil-agent-safety check --tool exec --params '{"command":"ls -la"}'
npx vigil-agent-safety policies

Links

• GitHub: https://github.com/hexitlabs/vigil
• npm: https://www.npmjs.com/package/vigil-agent-safety
• Docs: https://hexitlabs.com/vigil