csharp-refactor
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- PROMPT_INJECTION (HIGH): Vulnerable to Indirect Prompt Injection. The skill is designed to ingest and analyze external data (C# source code and project configuration files) and has the capability to modify the file system using 'Edit' and 'Write' tools. This creates a significant attack surface where instructions hidden in comments could be executed by the model.
- Ingestion points: The skill reads .cs and .csproj files from the repository context using Read, Glob, and Grep.
- Boundary markers: No delimiters or instructions are provided to distinguish between code and potentially malicious embedded natural language instructions.
- Capability inventory: The skill has 'Edit' and 'Write' permissions, allowing it to apply changes based on potentially poisoned input.
- Sanitization: There is no validation or sanitization of the ingested code content.
- COMMAND_EXECUTION (LOW): The execution guidelines suggest running 'dotnet test' to verify changes. While the YAML frontmatter does not explicitly list a command execution tool, any environment that supports this instruction implicitly risks executing malicious code defined within the project's test suite.
Recommendations
- AI detected serious security threats
Audit Metadata