The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/podcastfy_generate.py invokes several system-level commands and external utilities using subprocess.run, including ffmpeg, ffprobe, pip, and edge-tts.
[EXTERNAL_DOWNLOADS]: During its setup and execution phase, the skill automatically fetches and installs Python packages (podcastfy, playwright) from the public PyPI registry and downloads browser binaries via playwright install.
[REMOTE_CODE_EXECUTION]: The script dynamically constructs a Python code block and executes it using the python -c flag. This, combined with the runtime installation of external packages, introduces a risk of executing unverified code.
[PROMPT_INJECTION]: The skill retrieves and processes content from arbitrary user-provided URLs. This architecture is susceptible to indirect prompt injection attacks, where instructions hidden on a webpage could influence the behavior of the AI model. 1. Ingestion points: URLs provided as arguments to the podcastfy_generate.py script. 2. Boundary markers: None identified in the wrapper script to separate external content from instructions. 3. Capability inventory: The skill can perform network requests, write files, and execute subprocesses like ffmpeg. 4. Sanitization: No explicit sanitization or filtering of the fetched webpage content was observed in the provided wrapper script.

podcastfy-clawdbot