The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill explicitly instructs the agent to check for and use $HOME/youtube_cookies.txt. This file contains sensitive session tokens and authentication data for YouTube accounts. Accessing credential stores from the user's home directory is a high-risk pattern that exposes authentication data.
[COMMAND_EXECUTION]: The workflow relies on executing shell commands such as rg (ripgrep) and yt-dlp. These commands process data derived from user-provided URLs, which can be a vector for command injection if input is not strictly validated by the underlying tool or the agent.
[DATA_EXFILTRATION]: The skill accesses sensitive local file paths, including personal Obsidian vaults and authentication cookie files. While the access is intended for local processing, the ability to read sensitive files is a prerequisite for exfiltration.
[PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection because it processes untrusted external content (YouTube transcripts) without adequate safeguards.
Ingestion points: External YouTube transcripts retrieved via yt-dlp or MCP tools.
Boundary markers: Absent; there are no instructions to delimit or ignore embedded instructions within the transcript data.
Capability inventory: The skill can execute shell commands (rg, yt-dlp) and perform file-system writes to the user's Obsidian vault.
Sanitization: Basic character replacement is performed for filenames (safe_video_id), but no sanitization or validation is applied to the content of the transcripts used for summary generation.

youtube-processing