The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes system utilities including curl, ffmpeg, and ffprobe via the subprocess module.
Command arguments are carefully managed to avoid shell injection, particularly in the safe_download.py utility which uses specific security-hardened flags.
[EXTERNAL_DOWNLOADS]: The skill downloads podcast audio and RSS metadata from external web sources.
It features a dedicated safe_download.py module designed to mitigate Server-Side Request Forgery (SSRF) and DNS rebinding attacks.
It validates that resolved IP addresses are globally routable (not private or loopback) and uses DNS pinning (--resolve) during downloads.
Redirects are followed manually with re-validation of each hop to ensure the final destination is also a public resource.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content such as RSS XML feeds and webpage metadata (titles, audio links).
The implementation includes filename sanitization and uses structured data serialization (JSON) for Markdown frontmatter to prevent content from influencing the agent's context or damaging output integrity.
[DYNAMIC_EXECUTION]: The scripts use a custom helper function to dynamically import internal modules.
This mechanism is scoped to load only Python files from the skill's own scripts directory, preventing arbitrary code execution from outside sources.

podcast-transcribe