Back to skill

Security audit

Taste Skill

Security checks for vulnerabilities and agentic risk

Overview

This is an opinionated frontend design guide with disclosed external asset and package-use guidance, but no hidden persistence, credential access, or destructive behavior.

Use this skill when you want a strongly opinionated frontend design reviewer/builder. Before allowing install or scaffold commands, pin package versions and review generated dependency changes. Treat image generation and third-party image/CDN use as external sharing, especially for confidential brand briefs or unreleased product details.

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
SKILL.md:1006
Finding
Unpinned Package Retrieval and Execution Through npx## Vulnerability Details **File Location**: `SKILL.md`, lines 1006-1007 **Vulnerability Type**: Unpinned third-party package execution **Risk Level**: Medium ### Vulnerable Code ```bash # shadcn/ui (open code, owned components) npx shadcn@latest init npx shadcn@latest add button card badge separator input ``` A related recommendation also appears at line 99: ```markdown | Modern SaaS where you own the components | shadcn/ui (`npx shadcn@latest add ...`) | You own the code, easy to customise; never ship default state | ``` ### Technical Analysis The skill instructs the agent to invoke `npx` with the mutable `latest` distribution tag. If the package is not already present locally, `npx` can retrieve it from the configured npm registry and immediately execute its command-line entry point. Because `latest` does not identify an immutable, previously reviewed release, the effective code executed by this instruction can change after the skill itself has been audited. The instruction does not require an exact version, lockfile validation, package-integrity review, registry verification, or execution in a restricted environment. This creates a supply-chain trust boundary: security depends on the future state of the package, its maintainers, the npm account and registry configuration, and the dependencies resolved at execution time. The audit found no evidence that the referenced package is currently malicious; the vulnerability is the unsafe, unpinned execution pattern. ### Attack Path 1. An attacker compromises the package maintainer account, package release process, npm registry path, or one of the dependencies included in a future release. 2. The attacker publishes malicious or compromised code under the version referenced by the mutable `latest` tag. 3. An agent follows the skill and runs `npx shadcn@latest init` or the subsequent `add` command. 4. `npx` downloads and executes the newly resolved package code. 5. The malicious process runs with the operatin ...[truncated 1108 chars]
Remediation
## Remediation Suggestions 1. Replace the mutable `latest` tag with an exact, reviewed version: ```bash npx --yes shadcn@X.Y.Z init npx --yes shadcn@X.Y.Z add button card badge separator input ``` 2. Document the approved version in the skill and update it only after a security review. 3. Verify package provenance and registry configuration before execution: ```bash npm config get registry npm view shadcn@X.Y.Z version dist.integrity dist.tarball ``` 4. Download and inspect the package before running its entry point when operating in a sensitive environment: ```bash npm pack shadcn@X.Y.Z ``` 5. Preserve and review lockfile changes generated by the command. Use reproducible installation controls such as `npm ci` for subsequent installs. 6. Run scaffolding tools in a sandbox or disposable workspace with: - No production credentials in the environment. - Access limited to the target project directory. - Restricted outbound network access. - No administrative or root privileges. 7. Review all generated files and dependency changes before committing or executing the generated application. 8. Apply the same exact-version policy to every executable package command documented by the skill.
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (6)

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.

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.

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.

Context-Inappropriate Capability

Medium
Confidence
91% confidence
Finding
Mandating image-generation tools as a first step introduces an undeclared capability and can trigger tool use or external service interaction that the user did not request. In an agent setting, this can lead to unexpected data sharing of user briefs with image providers, cost-incurring actions, or execution paths outside the declared scope of a frontend design skill.

Context-Inappropriate Capability

Medium
Confidence
88% confidence
Finding
The skill explicitly directs the agent to fetch remote images, CDN-hosted logos, and other external assets as part of normal operation. That broadens the agent's effective network surface beyond pure code generation and can cause unreviewed external requests, privacy leakage through URL fetches, or inclusion of untrusted third-party resources into generated deliverables.

Natural-Language Policy Violations

Medium
Confidence
91% confidence
Finding
The text requires the agent to design for both modes from the start and to avoid shipping light-only or dark-only unless explicitly instructed otherwise. This forces a specific presentation preference on users rather than offering a user or project choice, which fits the language/locale policy category for imposed user settings without opt-in.

Static analysis

No suspicious patterns detected.