codeguard-r
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes R source files that could contain embedded malicious instructions aimed at manipulating the agent's behavior during linting or formatting tasks.
- Ingestion points: R source files (.r) in the project repository.
- Boundary markers: Relies on agent following a structured output template, but lacks technical delimiters for the processed code content.
- Capability inventory: Execution of shell commands via Rscript and file system write access for formatting operations.
- Sanitization: No specific sanitization or filtering of input file content is performed.
- [DYNAMIC_EXECUTION]: The skill utilizes Rscript -e to execute R code strings directly from the command line for linting and formatting.
- Evidence: Found in SKILL.md and toolchain.md as the primary method for invoking lintr.
- [COMMAND_EXECUTION]: The skill executes shell commands using Rscript to perform code analysis and formatting.
- Evidence: Execution of Rscript -e command patterns to run lintr and styler.
Audit Metadata