Back to skill

Security audit

sitemd plugin

Security checks for vulnerabilities and agentic risk

Overview

The skill is a real website-management tool, but its install flow and deployment/authentication powers need review before use.

Install only if you trust the sitemd publisher and are comfortable with npm postinstall fetching native code from GitHub Releases. Use a test project first, avoid running install as an administrator, verify deployment targets before using deploy or activate, and treat login URLs and SITEMD_TOKEN/API keys like credentials.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Findings (1)

T03 · Remote Payload Retrieval and Execution

Error
Location
sitemd/install.js:158
Finding

Unverified Remote Native Binary Download and Execution

Content
View full analysis
{ const follow = (u) => { https.get(u, { headers: { 'User-Agent': 'sitemd-install' } }, res => { if (res.statusCode >= 300 && res.statusCode < 400 && res.headers.location) { return follow(res.headers.location) } if (res.statusCode !== 200) { return reject(new Error(`Download failed: HTTP ${res.statusCode} ${u}`)) } const file = fs.createWriteStream(dest) res.pipe(file) file.on('finish', () => { file.close(); resolve() }) file.on('error', reject) }).on('error', reject) } follow(url) }) } ``` ```js const archive = `sitemd-${wanted}-${platform}-${arch}${ext}` const url = `${GITHUB_RELEASE_URL}/v${wanted}/${archive}` const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'sitemd-install-')) const archivePath = path.join(tmpDir, archive) const extractDir = path.join(tmpDir, 'extracted') try { console.log(` sitemd: downloading v${wanted} (${platform}-${arch})...`) await download(url, archivePath) fs.mkdirSync(extractDir, { recursive: true }) if (ext === '.zip') { // Windows 10 1803+ ships tar.exe which handles zip files execFileSync('tar', ['-xf', archivePath, '-C', extractDir], { stdio: 'ignore' }) } else { execFileSync('tar', ['-xzf', archivePath, '-C', extractDir], { stdio: 'ignore' }) } ``` ```js const extractedBinary = path.join(sourceRoot, archiveBinaryName) if (!fs.existsSync(extractedBinary)) { throw new Error(`Binary not found in archive: ${ ...[truncated 5178 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (14)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill description frames the package as a website-management helper, but the instructions direct the agent to run an installer that downloads and executes a binary, then launches it. This is dangerous because it hides supply-chain and code-execution behavior behind a higher-level description, making it easier for an agent or user to approve execution without understanding that untrusted code will be fetched and run.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding

The skill advertises environment-based authentication via SITEMD_TOKEN and operational capabilities that can affect external systems, but it does not declare an explicit tool scope or permissions boundary. In agent environments, missing scope declarations increase the chance that the skill receives broader access than intended, enabling misuse of secrets or privileged operations without clear guardrails.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill exposes high-impact operations such as deploy, delete, activate, and update, but the instructions do not require confirmation or warn that these actions can irreversibly change a live site. In an autonomous or semi-autonomous agent context, that omission materially raises the risk of accidental destructive changes or unintended production releases.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The authentication flow tells the agent to send a login URL through third-party messaging services without warning about interception, account mix-ups, or privacy leakage. Login links are effectively bearer secrets during their validity window, so distributing them over external channels can enable unauthorized account access if messages are exposed or sent to the wrong recipient.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SOUL.md (reported line 71)May include surrounding context.

md
## Boundaries

- Never deploy without confirmation from your owner
- Never share authentication tokens or API keys in messages
- Ask before overwriting existing pages
- If you're unsure about site identity or brand voice, ask rather than guess

Rp1

Medium
Category
MCP Rug Pull
Confidence
89% confidence
Finding

The manifest directs users to install or scaffold via npx @sitemd-cc/sitemd without pinning an exact version, which allows whatever package version is current at execution time to be fetched and run. Because this plugin also relies on a binary obtained via postinstall or a helper script, the supply-chain risk is increased: a compromised upstream release or dependency could lead to arbitrary code execution during installation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The description states that the required binary is auto-fetched via npm postinstall or installed via a helper script, but it does not clearly warn users that installing the plugin may download and execute additional code. Hidden or under-disclosed install-time execution is dangerous because users may treat the plugin as static metadata while it actually triggers network retrieval and local code execution.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The file header claims the script 'downloads the platform binary,' but the implementation also initializes project content and modifies files in the caller project. This mismatch is security-relevant because it can mislead reviewers and users about the true scope of install-time behavior, making unexpected filesystem changes easier to hide inside a dependency install flow.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The installer goes beyond fetching a binary: during npm postinstall it scaffolds a ./sitemd/ project and writes agent-related files into the caller's project root based on INIT_CWD. This creates unexpected side effects in the consuming repository, which is dangerous because install-time scripts run with the user's privileges and can silently alter project configuration and agent behavior without explicit consent.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill explicitly instructs the agent to transmit a login URL over third-party messaging channels without any warning about link sensitivity, recipient verification, or phishing/session-hijack risk. Magic-link authentication URLs are effectively bearer secrets during their validity window, so sending them through external channels can expose account access if intercepted, misdirected, or logged by the messaging platform.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The MCP tools list advertises actions that can modify content, remove pages, change configuration, deploy changes, permanently activate a site, and apply updates. The markdown does not provide any cautionary note about confirming these actions, reviewing impacts, or backing up data before irreversible changes.

Content

No source excerpt is available for this finding.