Back to skill

Security audit

Track houses for sale on a private map

Security checks for vulnerabilities and agentic risk

Overview

The skill largely matches its house-tracking purpose, but it should be reviewed because it can open arbitrary listing URLs in a browser without blocking local or private network destinations.

Install only if you are comfortable with a house-hunting helper that opens listing pages in Plow Browser and writes a local property map. Use trusted public listing URLs, avoid unknown or shortened links, and prefer a version that blocks localhost, private-network, link-local, and metadata addresses before browser navigation and after redirects.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/scrape.ts:82
Finding

Server-Side Request Forgery Through Unrestricted Listing URL Navigation

Content
View full analysis
{ await browser.goto(url); const deadline = Date.now() + timeoutMs; ``` ```ts async function main(): Promise { const url = process.argv[2]; if (!url || !/^https?:\/\//i.test(url)) { toolError("usage: node scrape.ts ''"); return; } // Read before opening the browser. This is the guard: a missing or damaged // store fails here with an actionable message, rather than after the scrape, // the geocode, and a photo download that has already written an image. // It also gives us what this property looked like before the refresh. const dir = resolveDataDir(); const before = readStore(dir); const { connect } = require('plow-browser'); let browser: any; try { browser = connect(process.env.CAMOUFOX_RPC_ENDPOINT); } catch (err) { if ((err as Error & { name?: string })?.name === 'BrowserNotConnected') { return toolError('browser unavailable: endpoint not provisioned — enable Plow Browser in Settings'); } throw err; } const { scraped, photoUrl } = await scrapeRendered(browser, url); ``` ### Technical Analysis The command accepts a user-controlled listing URL and verifies only that its textual form begins with `http://` or `https://`. It does not reject loopback, RFC1918 private, link-local, carrier-grade NAT, IPv6 local, or cloud instance metadata destinations before passing the URL to `browser.goto()`. The project already implements robust destination validation for hero-image downloads through `resolvePublicDestination()`, including private-address rejection and DNS-rebinding mitigation. That protection is ...[truncated 2746 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (14)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The declared description describes a full-featured real estate tracking skill with ingestion, persistence, browser access, and map rendering. The supplied code chunk contains only //# sourceMappingURL=leaflet.js.map in a vendored frontend file path. This does not demonstrate the declared primary purpose or capabilities. While Leaflet is commonly used for maps and could be a supporting dependency for rendering pins, this snippet alone is unrelated to most of the declared behavior and does not substantiate the skill description. Therefore, the description does not accurately represent what this code chunk actually does.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The declared description is about a real-estate property tracking/map skill with listing ingestion, storage under ~/Plow/properties, and property-management actions. The supplied code does none of that. Instead, it is a test enforcing an invariant about user-facing command strings in source/docs/UI, specifically checking that script names are invoked with 'node'. This is a materially different primary purpose and accesses different resources than described. Therefore the description does not accurately represent this code chunk.

Content

No source excerpt is available for this finding.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 128)May include surrounding context.

md
- **Only ever change the store through `scrape.ts` or `properties.ts`** —

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 129)May include surrounding context.

md
- **Only ever change the store through `scrape.ts` or `properties.ts`** —

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 128)May include surrounding context.

md
- **Only ever change the store through `scrape.ts` or `properties.ts`** —

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 129)May include surrounding context.

md
- **Only ever change the store through `scrape.ts` or `properties.ts`** —

Cloud Metadata Access

High
Category
Server-Side Request Forgery
Confidence
90% confidence
Finding

Code accesses a cloud instance metadata endpoint (e.g. 169.254.169.254). A single request can return temporary IAM credentials, making this a high-value SSRF target for credential theft.

Content

Scanner excerpt · scripts/scrape.test.ts (reported line 182)May include surrounding context.

ts
expected: true,
      addresses: [
        '127.0.0.1', '127.1.2.3', '0.0.0.0', '10.0.0.5', '172.16.9.9', '172.31.255.1',
        '192.168.1.1', '169.254.169.254', '100.64.0.1', '224.0.0.1',
        '::1', '::', 'fc00::1', 'fd12::3', 'fe80::1', '[::1]',
        // Spellings the first version of this guard let through — one address
        // has many texts, so these are judged by value.

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/scrape.test.ts (reported line 210)May include surrounding context.

ts
// pattern-matching the URL text.
  await assert.rejects(() => resolvePublicDestination('http://localhost/x.jpg'), /non-public address/);
  await assert.rejects(() => resolvePublicDestination('http://127.0.0.1:9222/x.jpg'), /non-public address/);
  await assert.rejects(() => resolvePublicDestination('file:///etc/passwd'), /refusing a file: image URL/);
});

test('a redirect target is vetted against the same rule as the first hop', async () => {

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/store.test.ts (reported line 247)May include surrounding context.

ts
// pattern-matching the URL text.
  await assert.rejects(() => resolvePublicDestination('http://localhost/x.jpg'), /non-public address/);
  await assert.rejects(() => resolvePublicDestination('http://127.0.0.1:9222/x.jpg'), /non-public address/);
  await assert.rejects(() => resolvePublicDestination('file:///etc/passwd'), /refusing a file: image URL/);
});

test('a redirect target is vetted against the same rule as the first hop', async () => {

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding

The skill instructs the agent to use shell commands, browser/network access, and filesystem writes, but the manifest declares no explicit tool scope or permissions boundaries. In practice this increases blast radius: a prompt injection in listing content, a malformed URL, or an operator mistake could cause unintended command execution or access beyond what the skill actually needs.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The invocation description is broad enough to trigger on many ordinary house-hunting messages, screenshots, and URLs, which can cause the agent to invoke a powerful skill unexpectedly. Because this skill performs browser lookups, shell commands, and writes to a host-mounted folder, accidental activation materially increases the chance of unintended data modification or exposure.

Content

No source excerpt is available for this finding.

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · LICENSE (reported line 12)May include surrounding context.

text
the Software, and to permit persons to whom the Software is furnished to do so.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR
COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
92% confidence
Finding

The document declares lang="en", which fixes the page language to English. Under the policy rule for natural-language constraints, this is a locale/language restriction with no visible opt-in or alternate language handling.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The code formats prices and square footage with toLocaleString('en-US'), which enforces a specific locale for displayed content. This is a natural-language/locale policy issue because users are not given a choice and the file contains no justification for restricting formatting to US English conventions.

Content

No source excerpt is available for this finding.