Back to skill

Security audit

File Launcher

Security checks for vulnerabilities and agentic risk

Overview

This skill does what it says, but it tells the assistant to launch any Windows-associated file or program without confirmation or safety limits.

Review this skill carefully before installing. It should only be used if you are comfortable with an assistant opening local files and programs on Windows; safer wording would require explicit confirmation, show the resolved path, and block executables, scripts, shortcuts, installers, and untrusted locations by default.

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
SKILL.md:17
Finding

Unrestricted File and Program Launch Without User Confirmation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 17–19
Vulnerability Type: Unsafe invocation of arbitrary files and programs
Risk Level: Medium

Vulnerable Code

markdown
- The assistant will use `Invoke-Item` to open files
- No need to wait for confirmation as Windows programs are UI-based
- Opens any file for which Windows has a default player or program to manage it

Technical Analysis

The skill instructs the agent to pass files to PowerShell's Invoke-Item without requiring user confirmation and without defining restrictions on file type, source, or location. On Windows, Invoke-Item does not merely display passive content: it invokes the operating system action associated with the supplied item.

If the target is an executable or another active file type, this behavior may launch code under the current user's security context. Documents and media can also invoke registered applications and expose their attack surface. The statement that confirmation is unnecessary because applications are UI-based is therefore not a valid security control.

The reviewed skill contains no path-validation procedure, trusted-directory restriction, file-extension allowlist, executable-content rejection, or confirmation boundary for active content.

Attack Path

  1. An attacker causes an executable, script-associated file, shortcut, or crafted active document to exist at a location accessible to the user.
  2. The attacker or misleading context prompts the agent to open or launch that item.
  3. The agent follows the skill's instruction not to request confirmation.
  4. The agent supplies the target to Invoke-Item.
  5. Windows invokes the target itself or its registered handler under the current user's privileges.
  6. Any resulting code or application behavior gains access to resources available to that user.

Impact Assessment

Successful exploitation can cause unintended local program execution or processing of malicious content. Result ...[truncated 324 chars]

Remediation
View remediation

Remediation Suggestions

  1. Require explicit user confirmation before launching executables, shortcuts, scripts, installers, macro-enabled documents, or other active content.
  2. Resolve the target to a canonical absolute path before invocation and verify that it remains within a user-approved or trusted directory.
  3. Use a strict allowlist of passive file types required by the skill's legitimate purpose.
  4. Reject executable and script-capable extensions by default, including .exe, .com, .bat, .cmd, .ps1, .msi, .scr, .lnk, and similar active formats.
  5. Validate the actual file type rather than relying exclusively on its extension.
  6. Clearly display the resolved path, file type, and application that will handle the item before seeking confirmation.
  7. Treat files from downloads, temporary directories, network shares, removable media, and other untrusted sources as high risk.
  8. Invoke the selected handler through a constrained implementation rather than accepting arbitrary command text or user-controlled PowerShell expressions.
  9. Log launch decisions and failures without recording sensitive file contents.
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill explicitly states that files will be opened with Invoke-Item and that no confirmation is needed before launch. Removing user confirmation for a capability that can open arbitrary files or associated applications increases the risk of unintended execution, opening dangerous file types, or triggering handlers for URLs, scripts, shortcuts, and other risky shell-associated content.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The trigger phrases and usage examples are extremely generic (e.g. 'Open the image', 'Launch the document') and overlap with ordinary assistant requests. That broad matching can cause the skill to activate unexpectedly and perform local file or program launching when the user did not explicitly intend to invoke a powerful file-opening capability.

Content

No source excerpt is available for this finding.