The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute generated R scripts and supports background processing for long-running simulations.
[DYNAMIC_EXECUTION]: The skill generates R scripts based on user input and templates, saving them to the local scripts/R/ directory before execution. This behavior is consistent with the primary purpose of the skill.
[INDIRECT_PROMPT_INJECTION]: The skill accepts user-provided descriptions of estimators and data generating processes, which are interpolated into generated code.
Ingestion points: The $ARGUMENTS parameter in SKILL.md captures user input and pointers to external scripts or papers.
Boundary markers: The workflow mandates a 'Pre-Flight Report' (Phase 0) to define estimands, assumptions, and DGP parameters before code is written.
Capability inventory: The skill uses Bash, Write, Edit, Agent, Task, and Monitor tools (SKILL.md).
Sanitization: The 'Phase 6: Save & Review' step requires running a sim-reviewer agent to audit generated scripts for errors and safety issues.
[METADATA_POISONING]: The skill metadata lists the author as 'Claude Code Academic Workflow', which does not match the provided context author 'pedrohcgs'. This appears to be a descriptive label for the workflow template rather than an attempt at malicious impersonation.

simulation-study