The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The download_videos_from_manifest_with_ytdlp.mjs script fetches external video assets based on URLs provided in a JSON manifest using the well-known yt-dlp utility.
[COMMAND_EXECUTION]: The skill performs subprocess execution using child_process.spawn to run python3 -m yt_dlp. This is used to programmatically download and verify media files as part of the production workflow.
[CREDENTIALS_UNSAFE]: The postplus_cli_config.mjs module accesses the local PostPlus configuration directory to read config.json. It retrieves session tokens, account identifiers, and API base URLs to authenticate the agent with the vendor's cloud platform.
[PROMPT_INJECTION]: The skill processes research artifacts, such as master tables and comment analyses, which represents an indirect prompt injection surface.
Ingestion points: SKILL.md (research artifacts) and download_videos_from_manifest_with_ytdlp.mjs (manifest file).
Boundary markers: The skill includes instructional constraints in SKILL.md (the 'Fact Rule') requiring fact-grounding, but does not implement programmatic delimiters in its data-handling scripts.
Capability inventory: The skill has the ability to execute shell commands (yt-dlp), perform file system operations (saving videos and reports), and make network requests via vendor-provided libraries.
Sanitization: Subprocess calls in the downloader script use argument arrays, which prevents shell command injection from malicious manifest data.

benchmark-to-brief