The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides run_command and run_do_file tools that enable the execution of arbitrary Stata syntax and script files on the host machine. This is a core functionality but represents an execution surface that could be exploited.
[DATA_EXFILTRATION]: The read_log tool accepts a file path and returns its content. Without explicit server-side restrictions to specific directories, this capability could be used to read sensitive local files.
[EXTERNAL_DOWNLOADS]: The load_data tool supports fetching datasets from arbitrary URLs, which may involve downloading content from untrusted external sources.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes external datasets which could contain instructions intended to manipulate the agent's behavior.
Ingestion points: The load_data tool in SKILL.md can ingest data from remote URLs and local file paths.
Boundary markers: There are no instructions provided to the agent to ignore or delimit embedded instructions within the ingested data.
Capability inventory: The skill includes code execution (run_command), script execution (run_do_file), and file reading (read_log).
Sanitization: No sanitization or validation logic is described for the data processed by the tools.

stata-mcp