Back to skill

Security audit

Mortgage Rate Lookup

Security checks for vulnerabilities and agentic risk

Overview

This mortgage-rate skill mostly does what it says, but it includes under-disclosed browser access, broad third-party ZIP submission, and inconsistent dependency installation details that users should review before installing.

Review this skill before installing. It will contact multiple mortgage and lender websites, may submit your ZIP code to those sites, installs browser automation components, and keeps local rate history. The local OpenClaw browser CDP helper and dependency installation mismatch should be fixed or clearly opt-in before routine use.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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)

T08 · Insecure Dependencies

Warning
Location
requirements.txt:1
Finding

Unpinned and Inconsistently Declared Third-Party Dependency

Content
View full analysis

Vulnerability Details

File Location: requirements.txt:1, SKILL.md:14-17, and SKILL.md:35-36
Vulnerability Type: Supply-chain risk through mutable, unverified third-party dependencies
Risk Level: Medium

Vulnerable Code

requirements.txt:1:

text
patchright>=1.58.0

SKILL.md:14-17:

yaml
install:
  - kind: node
    package: patchright
    bins: []

SKILL.md:35-36:

bash
pip install patchright
python3 -m patchright install chromium

Technical Analysis

The project permits any Patchright release at or above version 1.58.0 and does not provide integrity hashes for the package or downloaded Chromium artifacts. The documented installation command is even less restrictive because it installs the latest available Python release.

In addition, the Skill metadata declares Patchright as a Node package, while the implementation imports the Python package through patchright.async_api. This inconsistency unnecessarily involves two package ecosystems and may cause an automated installer to retrieve a component that is not required by the program.

Because the permitted dependency and browser artifacts can change after this Skill has been reviewed, the code ultimately executed on a user's system is not fully represented by the audited project files. This creates a supply-chain exposure if a future package release, package registry account, distribution channel, or downloaded browser artifact is compromised.

No evidence was found that the current project intentionally introduces a malicious package. The issue is the absence of deterministic version and integrity controls.

Attack Path

  1. An attacker compromises a future Patchright release, a relevant package-registry account, or its artifact distribution infrastructure.
  2. The compromised version remains compatible with the open-ended patchright>=1.58.0 constraint or is returned by pip install patchright.
  3. A user or automated Skill installer follows the ...[truncated 1116 chars]
Remediation
View remediation

Remediation Suggestions

  1. Replace the open-ended requirement with an exact, reviewed Python package version:
text
patchright==1.58.0
  1. Generate a lock file containing cryptographic hashes for Patchright and all transitive Python dependencies, and install with hash verification, such as:
bash
pip install --require-hashes -r requirements.lock
  1. Remove the Node-based Patchright installer declaration from SKILL.md unless a separately reviewed Node dependency is genuinely required. Declare only the Python installation path used by mortgage_rate_report.py.

  2. Pin the Chromium revision used by Patchright where supported. Verify downloaded browser artifacts using vendor-provided checksums or signatures before execution.

  3. Perform dependency installation in an isolated virtual environment or container under a non-privileged account.

  4. Use automated dependency monitoring, but update pinned versions only after reviewing release provenance, checksums, dependency changes, and security advisories.

  5. Keep installation documentation and Skill metadata synchronized so automated installers cannot select a different ecosystem or artifact from the one audited.

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (9)

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The documented purpose does not fully match the detected behavior: it omits important operational details such as network access, file writes, and a fallback connection to a local browser over CDP/OpenClaw. That mismatch is dangerous because hidden or under-disclosed capabilities can be abused to access local browsing context or exfiltrate data under the cover of a benign mortgage-rate lookup function.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill advertises no explicit tool scope, yet its documented behavior clearly requires network access and local file writes (config.json, history tracking, browser installation/use). Undeclared capabilities weaken review and enforcement boundaries, making it easier for a consumer or platform to run the skill with broader access than expected.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The module docstring and CLI help repeatedly state the tool compares 13 lenders plus benchmarks, but BROWSER_SOURCES contains only 9 lenders and both BROWSER_ASSISTED and HEADED_SOURCES are empty. Although a Chase-specific fallback exists later, it is never invoked from main execution, so the actual implemented coverage is materially narrower than described.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The script automatically submits a user ZIP code to multiple third-party lender sites during scraping without explicit notice at the point of use. ZIP code is location data, and broadcasting it across many external services can create an avoidable privacy leak and user-surprise issue, especially in a skill context where actions may be run on the user's behalf.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

The comments assert a completed migration state ('all 10 are automated' and no additional headed sources), but the active list has 9 lenders and a separate Chase CDP scraper remains in the file. This is not merely incomplete documentation; it actively misstates the implemented state of the code.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

The browser automation explicitly sets locale="en-US", which hard-codes a specific language/locale behavior. There is no opt-in, configuration option, or documented justification that this skill is intended only for an English/US-specific context.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The script persists rate history locally without explicit disclosure to the user. While the stored file contains rates rather than the ZIP code itself, local persistence still creates a privacy and transparency issue because users may not expect ongoing storage of activity-derived data.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
98% confidence
Finding

The comment documents total_lenders as representing 10 automated lenders, yet len(BROWSER_SOURCES) is currently 9. This contradicts the actual configured behavior and can mislead maintainers about the scope of the produced report.

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
94% confidence
Finding

The dependency is only lower-bounded and not pinned to an exact version, which makes builds non-reproducible and allows future releases of the package to be pulled in automatically. If an upstream release is compromised, malicious, or simply introduces insecure behavior, this skill could begin executing unreviewed code during installation or runtime.

Content

Scanner excerpt · requirements.txt (reported line 1)May include surrounding context.

text
patchright>=1.58.0