The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from the web that could contain malicious instructions.
Ingestion points: Data enters the context via WebSearch, WebFetch, and the mcp__claude-in-chrome__* browser tools as described in the Source Discovery and Content Gathering steps in SKILL.md.
Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands within the gathered content.
Capability inventory: The skill has access to the Bash tool (used for yt-dlp execution) and Write tool for file system operations as defined in the frontmatter and Step 3 of SKILL.md.
Sanitization: There is no evidence of sanitization or filtering of the content retrieved from external URLs before it is processed by the agent.
[COMMAND_EXECUTION]: The skill executes shell commands using yt-dlp via the Bash tool to download video segments. While these commands are aligned with the skill's purpose, they represent a potential injection vector if the {VIDEO_URL} parameter is not properly sanitized by the underlying platform.
[EXTERNAL_DOWNLOADS]: The skill performs automated downloads of media files from external URLs using yt-dlp. While expected for video research, this involves interacting with potentially untrusted remote servers.

video-research