The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts scripts/evaluate.py and scripts/dive_club.py use the subprocess.run() function to execute shell commands. This is used to compile a bundled Swift transcription helper and to invoke a command-line interface tool for model-based evaluation.
[EXTERNAL_DOWNLOADS]: In scripts/dive_club.py, the skill utilizes urllib.request.urlopen to fetch an RSS feed and download audio files from media.rss.com. These downloads are part of a research workflow designed to gather and transcribe podcast-based evidence.
[DYNAMIC_EXECUTION]: The skill performs runtime compilation of a Swift source file (scripts/transcribe.swift) into an executable binary using the swiftc compiler. This enables local, on-device audio transcription on macOS systems.
[INDIRECT_PROMPT_INJECTION]: The scripts/evaluate.py script generates prompts for model execution that incorporate data from external input.json files, such as portfolio requests and evidence descriptions. Without explicit boundary markers or strict sanitization of this content, instructions embedded within the portfolio data could potentially influence model behavior, though the script is intended for use in a controlled evaluation environment.

hiring-manager-bar