The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The SKILL.md file contains instructions for the agent to clone an external GitHub repository (https://github.com/Learning-Bayesian-Statistics/baygent-skills.git) if the dependency bayesian-workflow skill is not detected locally.
[REMOTE_CODE_EXECUTION]: The skill automates the installation of remote content by copying files from the cloned repository directly into the agent's skill directory (~/.claude/skills/). This introduces new instruction sets and logic into the agent's execution environment from an external source.
[COMMAND_EXECUTION]: The skill provides explicit shell commands for the agent to execute, including git clone, mkdir, and cp -r. These commands facilitate environment modification and network access to fetch external resources.
[DATA_EXPOSURE]: The skill possesses an indirect prompt injection surface through the ingestion of external datasets (e.g., CSVs or DataFrames) for causal analysis.
Ingestion points: Dataframes passed to model constructors such as CausalModel in references/dags-and-identification.md and SyntheticControl in references/quasi-experiments.md.
Boundary markers: Absent; there are no specific instructions to the agent to treat data content as untrusted or to ignore instructions embedded within the data.
Capability inventory: The skill environment allows for file system writes (plotting and reporting), shell command execution, and arbitrary Python execution.
Sanitization: Absent; the workflow does not include validation or sanitization of data column contents before processing.

causal-inference