The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill generates R scripts from templates based on user-supplied parameters and executes them. This includes the definition of Data Generating Processes (DGP) and estimators that are written to scripts in scripts/R/ and subsequently executed. The use of saveRDS() for results also introduces a minor risk, as RDS files can contain serialized objects that execute code upon loading.
[COMMAND_EXECUTION]: The skill uses the Bash tool to run the generated R code. It specifically describes how to background these processes using run_in_background: true, which allows for long-running code execution in the shell environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided descriptions of statistical models and potentially external papers/scripts to define simulation logic, creating a surface for indirect instructions to influence the generated code.
Ingestion points: User input provided via the $ARGUMENTS variable and references to external scripts or papers mentioned in the prompt.
Boundary markers: There are no explicit delimiters or boundary markers used when interpolating the user's simulation design into the agent's instructions.
Capability inventory: The skill is granted access to high-privilege tools including Write, Edit, Bash, and Agent (allowing it to delegate tasks).
Sanitization: The workflow incorporates multi-phase validation, including a mandatory 'Pre-Flight Report' (Phase 0) to verify assumptions and a 'sim-reviewer' agent check (Phase 6) to audit the generated code for logical and security errors before presentation.

simulation-study