find-untested-sources
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- Command Execution: The skill facilitates analysis by instructing the agent to execute provided local scripts (
scripts/Find-UntestedSources.csandscripts/find_untested_sources.py). This is a standard approach for agent skills that provide complex utility beyond simple prompting. - External Dependencies: The tool relies on legitimate external libraries for static analysis, including
Microsoft.CodeAnalysis.CSharpfor the Roslyn-based C# engine andtree-sitter-language-packfor the polyglot engine. These are well-established packages in the development community. - Indirect Prompt Injection Surface: The skill processes source code files from the repository to extract symbols and identifiers for pairing.
- Ingestion points: Source files are read during the walk phase in
scripts/Find-UntestedSources.csandscripts/find_untested_sources.py. - Boundary markers: The skill uses a structured JSON output schema to return results to the agent, providing clear separation between data and instructions.
- Capability inventory: The agent reports findings to the user; the scripts themselves only perform read-only static analysis and do not write to the file system or perform network operations.
- Sanitization: The scripts utilize formal syntax parsers (Roslyn and tree-sitter) which ensure only valid code identifiers are extracted, limiting the ability of arbitrary text in source files to influence the analysis results.
Audit Metadata