The Agent Skills Directory

[COMMAND_EXECUTION]: The skill constructs shell commands by directly interpolating user input into command strings within the bash execution environment. Specifically, the commands tavily-search "<query>" and tavily-crawl "<url>" use variables that originate from user input. Although wrapped in double quotes, this pattern can be vulnerable to command injection through shell features like command substitution (e.g., backticks or $()) if the underlying execution engine does not properly sanitize these inputs.
[PROMPT_INJECTION]: The skill provides a surface for indirect prompt injection by fetching and interpreting content from external, untrusted web sources.
Ingestion points: Search results (titles, snippets, AI answers) from tavily-search and full page content from tavily-crawl (found in SKILL.md).
Boundary markers: Absent. The instructions do not specify any delimiters or "ignore previous instructions" guards when the agent processes the fetched external content.
Capability inventory: The agent has the capability to execute shell commands (bash) and summarize data, which could be subverted by malicious instructions embedded in a crawled webpage.
Sanitization: No sanitization, escaping, or validation of the external content is performed before the agent is instructed to summarize and interpret the results for the user.

tavily-search