Randomization Gen

Generate block randomization lists for RCTs

This skill appears to do what it claims (generate block and stratified randomization lists) and does not request secrets or network access. Before installing or running it, consider: 1) Run it in a sandbox or restricted workspace — the script writes whichever filename you pass and does not validate paths, so it could overwrite files outside the intended directory if given a path like ../../something. 2) Validate user-supplied output paths yourself (or modify the script) to prevent directory traversal and to ensure you have permission to write. 3) Edge-case behavior: stratified_randomization uses integer division (n_subjects // len(strata)) and will drop any remainder subjects (they won't be allocated); consider whether this is acceptable or modify the logic to distribute remainders. 4) The exporter will fail if the schedule is empty (schedule[0].keys() will raise), so add checks for empty schedules if needed. 5) If you plan to run this on sensitive systems, confirm the runtime environment and filesystem permissions. These issues look like sloppy/omitted validation rather than intentional misdirection, but you should review or patch the code before using it in a production or high-stakes clinical environment.

Type: OpenClaw Skill Name: randomization-gen Version: 1.0.0 The skill bundle is a legitimate tool for generating clinical trial randomization lists. The Python script (scripts/main.py) implements standard block and stratified randomization logic using built-in libraries (random, csv, argparse) and contains no hidden network calls, file exfiltration, or malicious execution patterns. The documentation (SKILL.md) accurately reflects the code's functionality and does not contain any prompt-injection attempts.