The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a bash command using a search query: echo "<検索クエリ>" | gemini -p "...". This template is vulnerable to shell command injection. If the search query (which originates from user intent or conversation context) includes characters like backticks, semicolons, or command substitutions, it could allow an attacker to execute arbitrary code on the underlying system.
[INDIRECT_PROMPT_INJECTION]: The skill retrieves and processes data from external web sources via the gemini CLI tool.
Ingestion points: The output of the gemini CLI command (Google search results) in SKILL.md.
Boundary markers: No delimiters or warnings are used to instruct the agent to ignore instructions embedded in the search results.
Capability inventory: Bash execution is active as part of the primary skill workflow in SKILL.md.
Sanitization: No sanitization, escaping, or validation of the external content is performed before processing.

gemini-search