The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses child_process.spawn and execFileSync to execute local tools including yt-dlp and ffmpeg for downloading and processing TikTok media content. These operations are part of the core functionality for media research and benchmarking.
[EXTERNAL_DOWNLOADS]: The skill downloads media files from TikTok platforms via yt-dlp based on URLs provided in collection manifests. This is a documented feature for asset persistence and research.
[CREDENTIALS_UNSAFE]: The skill manages API configuration and session tokens for the vendor's platform within local workspace directories (.postplus/config.json). This follows standard practice for managed tool configuration and uses file-level access controls (0o600) for sensitive data.
[PROMPT_INJECTION]: The skill ingests untrusted TikTok data (such as user comments and profile signatures) which presents a surface for indirect prompt injection.
Ingestion points: Data is processed through scripts/analyze_tiktok_comments.mjs and scripts/analyze_tiktok_dataset.mjs.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are used when summarizing this content for the agent.
Capability inventory: The skill possesses capabilities for local command execution (yt-dlp, ffmpeg) and file system operations within the workspace.
Sanitization: The scripts perform tokenization and URL removal, which provides basic filtering against simple injection vectors in user-generated content.

tiktok-research