The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of R scripts and system-level operations.
Evidence: Usage of source("script.R") to run external R files.
Evidence: Capabilities to build and initialize packages using devtools::create().
Evidence: Execution of web applications via shiny::runApp().
[EXTERNAL_DOWNLOADS]: The skill documentation describes mechanisms for downloading and installing external software dependencies at runtime.
Evidence: Instructions to use install.packages() for fetching libraries from the Comprehensive R Archive Network (CRAN).
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection through the processing of external data sources.
Ingestion points: The skill reads external data via read_csv("data.csv") and configuration files via yaml.load_file("config.yml").
Boundary markers: No explicit instructions or delimiters are provided to the agent to disregard instructions embedded within these data files.
Capability inventory: The skill possesses high-impact capabilities including network requests (httr::RETRY), package installation (install.packages), and environment variable modification (Sys.setenv).
Sanitization: There is no mention of sanitizing or validating the content of ingested CSV or YAML files before they influence program logic or are passed to other functions.
[DATA_EXFILTRATION]: The skill handles sensitive credentials and possesses network capabilities.
Evidence: Instructions for managing API keys via environment variables and YAML configuration files using placeholders like $MY_API_KEY.
Evidence: Network connectivity is supported through the httr package for making web requests.

coding-r