The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the public web to generate deduplicated lists and tables.
Ingestion points: Processes external data from web pages and list roundups retrieved via the tavily-search tool.
Boundary markers: The instructions do not define delimiters or explicit "ignore" warnings to prevent the agent from obeying instructions embedded in the external web content.
Capability inventory: The skill utilizes Bash for the tvly tool, performs file writes to the local lists/ directory, and can execute npx commands.
Sanitization: There is no evidence of sanitization or structural validation performed on the external data before it is integrated into the agent's context or written to markdown files.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for the agent to install missing dependencies from the vendor's repository.
The instructions direct the agent to run npx skills add tavily-ai/skills to retrieve related tools such as tavily-search, tavily-entity, and tavily-research.
[COMMAND_EXECUTION]: The skill relies on shell command execution for installation and tool usage.
The configuration allows the use of the tvly tool through the Bash tool.
The instructions explicitly command the use of npx for dynamic package installation.

tavily-list