The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a Python script (scripts/prepare_citations.py) to process files. Analysis of the source code confirms it only uses standard libraries (argparse, json, re, pathlib, etc.) and does not invoke any shell commands or external processes using subprocess or os.system.
[DATA_EXFILTRATION]: The script reads input from local files (markdown and JSON) and writes the results back to a local file. There are no network-related imports (e.g., requests, urllib.request, socket) or operations that would allow data to be transmitted to a remote server.
[REMOTE_CODE_EXECUTION]: There is no evidence of dynamic code execution (eval, exec) or remote script downloads. The script only performs string manipulation and regex-based transformations on the provided text data.
[CREDENTIALS_UNSAFE]: No hardcoded credentials, API keys, or secrets were found in the markdown instructions or the Python script. It follows safe practices by processing data provided via command-line arguments.
[SAFE]: The skill is a well-documented utility tool. Its internal logic, such as the repair_claim_id_leakage function, is designed to fix specific data consistency issues from previous agent steps, and it uses standard sanitization patterns like _normalize_url and slugify (which correctly handles CJK characters).

sn-prepare-citations