Back to skill

Security audit

Find Skills.Local.Backup

Security checks for vulnerabilities and agentic risk

Overview

This skill is coherent, but it should be reviewed because it directs agents to run a live npm CLI and globally install third-party skills while skipping confirmation prompts.

Install only if you are comfortable letting an agent search for and install third-party skills. Before running the documented commands, prefer a pinned and reviewed Skills CLI version, inspect the skill source, avoid `-y`, and use global installation only when you intentionally want the skill available across sessions.

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:24
Finding

Unpinned Third-Party CLI Execution and Unscreened Global Skill Installation

Content
View full analysis
` - Install a skill from GitHub or other sources - `npx skills check` - Check for skill updates - `npx skills update` - Update all installed skills ``` ```markdown Run the find command with a relevant query: ```bash npx skills find [query] ``` ``` ```markdown If the user wants to proceed, you can install the skill for them: ```bash npx skills add -g -y ``` The `-g` flag installs globally (user-level) and `-y` skips confirmation prompts. ``` ### Technical Analysis The Skill instructs the Agent to execute `npx skills` without pinning the npm package to an exact, reviewed version or enforcing an integrity hash. Depending on local npm behavior and cache state, `npx` can retrieve and execute a mutable package release from the npm registry. The workflow subsequently allows Skills to be selected from search results and installed from GitHub or other sources. It does not require verification of repository ownership, source inspection, an immutable commit reference, or package integrity before installation. The recommended `-g -y` options increase the risk because they install the selected Skill at user scope and suppress the normal confirmation prompt. This is a supply-chain weakness rather than evidence that the currently documented `skills` package or any particular repository is malicious. Exploitation depends on compromise or malicious control of the CLI package, registry/search results, or a selected Skill repository. ### Attack Path 1. An attacker publishes, takes control of, or compromises a relevant npm package release, Skill repository, or source surfaced by the search mechanism. 2. The Agent fo ...[truncated 1209 chars]
Remediation
View remediation
`, and enforce package-lock or integrity verification where supported. 2. Do not automatically trust search results. Restrict installations to an allowlist of reviewed owners, repositories, and Skill identifiers. 3. Pin installed Skills to immutable commit hashes or cryptographically verified releases rather than mutable branches or tags. 4. Require source review before installation, including inspection of Skill instructions, scripts, hooks, dependencies, and requested permissions. 5. Remove `-y` so users receive an explicit confirmation prompt before installation. 6. Default to project-local or isolated installation instead of `-g`. Require separate, informed consent when global installation is genuinely necessary. 7. Present the exact source, version or commit, installation scope, and security implications to the user before executing the command. 8. Run third-party package discovery and installation with least privilege, preferably in a sandbox without access to sensitive credentials or user files. 9. Add a post-installation verification process that confirms installed files match the reviewed source and reports unexpected executable components or hooks. ]]>
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (17)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The metadata description positions the skill to activate for very broad requests like 'how do I do X' or general interest in extending capabilities. Overbroad activation increases the chance that the agent invokes a skill that recommends external package discovery or installation in situations where the user did not clearly ask for that, creating unnecessary exposure to supply-chain and system-modification actions.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The 'When to Use This Skill' section uses ambiguous triggers such as general capability questions or vague interest in help with a domain. In context, this is dangerous because the skill leads to executing external CLIs and potentially installing extensions, so ambiguous activation can turn ordinary help requests into package discovery and install flows without clear user intent.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

The skill repeatedly instructs use of npx skills without pinning a specific package version, which means execution depends on whatever version the registry serves at runtime. Because this skill is specifically about discovering and installing third-party skills, the unpinned package creates a supply-chain risk and could lead to execution of changed or compromised code.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

This reference invokes npx skills without an exact version, so the command may fetch and run a different package version over time. In a skill that guides users toward package discovery and installation, that increases exposure to registry compromise, malicious updates, or breaking changes becoming code-execution paths.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

The command example uses npx skills without version pinning, allowing arbitrary future package revisions to be executed. Since the command is presented as a normal workflow step, users or agents may run it automatically, amplifying the impact of any package compromise.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

This unpinned npx skills invocation introduces a supply-chain execution risk because it relies on the current latest package state rather than a reviewed version. In context, the skill teaches repeated operational use of the CLI, so the risk is systematic rather than incidental.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

Referencing npx skills without pinning allows the executed code to change outside the skill author's control. Because the command is tied to installing more code-like extensions, this creates a chain of trust issue that can expose the system to malicious or unexpected package behavior.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The search command is shown as npx skills find [query] with no version pin, meaning a fresh or updated package may be executed whenever the skill is followed. Even though this is a search action, it still requires running remote package code and therefore carries code-execution and supply-chain risk.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

This example instructs an unpinned npx skills find react performance command, which could execute a compromised or altered package version. Because examples are likely to be copied directly by users or automation, the risk is practical and not merely theoretical.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

The skill includes a concrete npx skills find pr review example without version pinning, creating the same supply-chain exposure as other invocations. The surrounding context normalizes executing the CLI as part of routine assistance, which increases the chance of agent-driven or user-driven execution.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

This example again runs an unpinned package via npx, allowing unreviewed upstream changes to alter what code is executed. Since the skill is intended to facilitate discovery and later installation of more extensions, it compounds trust in external code sources.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
91% confidence
Finding

The output example includes an install command using npx skills add <owner/repo@skill> without pinning the CLI version. While the package argument is partly specified, the CLI itself remains mutable and can change behavior or become compromised, which is risky in an installation workflow.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

This user-facing install example uses unpinned npx skills add, exposing users to running whatever package version is current at that moment. Because installation actions modify the agent's capability set and may fetch additional external content, the unpinned CLI materially increases supply-chain risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill tells the agent to offer installation using -g -y, which implies persistent global modification and suppresses confirmation prompts, but it does not require warning the user first. This can lead to silent or insufficiently informed system changes, especially harmful in an agent context where users may not realize the scope of installation side effects.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The command npx skills add <owner/repo@skill> -g -y combines unpinned remote package execution with global installation and confirmation skipping. This is more dangerous than the search examples because it can silently make persistent system-level changes while relying on a mutable upstream package.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
90% confidence
Finding

The instruction to create a skill with npx skills init is unpinned, so it still executes a mutable remote package version. Although the action is less directly dangerous than silent install, it retains the same supply-chain exposure and could scaffold malicious or altered content if upstream is compromised.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
90% confidence
Finding

This final npx skills init my-xyz-skill example again executes an unpinned remote package. In aggregate with the rest of the skill, it demonstrates a pattern of trusting live package resolution, which is unsafe for agent-directed operational guidance.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.