The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands via the Bash tool to execute tvly for research tasks and wc to monitor the length of generated reports. These operations are within the expected functional scope.
[EXTERNAL_DOWNLOADS]: The instructions include a setup step to install additional skills from the vendor's official repository (tavily-ai/skills) using the npx skills add command. As these originate from a trusted organization, they are considered safe vendor resources.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external web sources such as news articles, social media profiles, and company documentation.
Ingestion points: Data enters the agent context through the outputs of tavily-search and tavily-entity tools (SKILL.md).
Boundary markers: The skill does not define explicit delimiters for external content but includes a 'Quality bar' and specific verification instructions to match quotes against extracted text.
Capability inventory: The agent has the capability to execute shell commands (Bash) and perform file-system writes to save reports (SKILL.md).
Sanitization: The skill mitigates risks by instructing the agent to focus on specific, verifiable data points (dates, titles, and direct quotes) rather than executing instructions found within the data.

tavily-meeting-prep