The Agent Skills Directory

[SAFE]: The skill is purely instructional, guiding the agent through a manual code review process. It focuses on identifying logical errors and best practices in research scripts.
[PROMPT_INJECTION]: The skill exhibits an Indirect Prompt Injection surface because it is designed to ingest and process untrusted external data (Stata/R/Python source code files).
Ingestion points: Reads files from paths provided in $ARGUMENTS or searched project directories (e.g., ~/Dropbox/Github/).
Boundary markers: Absent; the instructions do not define specific delimiters or instructions to ignore embedded commands within the code files.
Capability inventory: The skill uses file reading to analyze code and file writing to generate markdown reports and optional "corrected" versions of scripts.
Sanitization: No sanitization or filtering is performed on the content of the scripts being reviewed.
[DATA_EXFILTRATION]: While the skill reads local files, it does not contain any instructions or patterns for sending data to external network destinations. Reports are saved locally in the project directory.

code-review