The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation includes code snippets for installing common statistical packages from CRAN and a specific academic package ('HonestDiD') from a researcher's GitHub repository ('asheshrambachan/HonestDiD'). These are standard procedures for quantitative researchers to access specialized statistical tools.
[COMMAND_EXECUTION]: The skill instructions direct the agent to use shell commands for project management ('git add/commit') and to execute R scripts ('Rscript'). These actions are consistent with the skill's core functionality of managing and running a statistical analysis project.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to read and process external data files provided by the user (e.g., CSV, Stata, or RDS files). This creates an inherent surface for indirect prompt injection where malicious content within the data could theoretically attempt to influence the agent's behavior during data inspection phases.
Ingestion points: The skill reads external files from the 'data/raw/' directory during the Phase 1 Data Familiarization process.
Boundary markers: No explicit delimiters or warnings to ignore embedded instructions are used when the agent processes data summaries.
Capability inventory: The skill can execute shell commands via 'Rscript' and 'git' as specified in the SKILL.md execution method.
Sanitization: The skill relies on standard R data loading functions (like 'read.csv' or 'haven::read_dta'), which do not execute data as instructions.

r-analyst