codex-windows-fast-patch-skill
Fail
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: HIGHPRIVILEGE_ESCALATIONDYNAMIC_EXECUTIONCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill provides instructions and workflows for creating a self-signed certificate and importing it into the
Cert:\LocalMachine\Root(Trusted Root Certification Authorities) store. This allows any code signed by that developer certificate to be trusted by the system, bypassing standard signature verification for all applications. - [DYNAMIC_EXECUTION]: The skill implements a workflow to extract an MSIX package, unpack its internal
app.asararchive, and rewrite JavaScript files to patch feature toggles, request parameters, and locale settings. This constitutes dynamic modification of application executable logic at runtime. - [COMMAND_EXECUTION]: All provided PowerShell commands use the
-ExecutionPolicy Bypassflag, which explicitly disables local security protections for script execution. The workflows also involve significant system-level operations such asGet-AppxPackage,makeappx.exe, andsigntool.exeto manipulate installed software. - [INDIRECT_PROMPT_INJECTION]: The skill identifies issues based on user-provided error reports and log messages (e.g., "Computer Use shows missing-helper-path error").
- Ingestion points: User prompts describing application failures and logs provided by the user (SKILL.md).
- Boundary markers: None identified in the skill instructions.
- Capability inventory: Extensive capabilities including local file system writes, application re-installation (MSIX), and network operations via PowerShell (SKILL.md).
- Sanitization: No evidence of input validation or sanitization before interpreting user reports to trigger patching workflows.
- [EXTERNAL_DOWNLOADS]: The skill fetches updates for its own logic from GitHub via
update-skill-from-github.ps1and usesnpx asarto download and execute the ASAR utility for archive manipulation.
Recommendations
- AI detected serious security threats
Audit Metadata