The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/search.sh script is vulnerable to shell command injection. The user-provided query is stored in the ESCAPED_QUERY variable and then interpolated into a curl command string wrapped in double quotes. Because the script fails to escape shell-sensitive characters such as backticks (`) and dollar signs ($), any shell code contained within these characters in the query will be executed by the shell during the execution of the curl command.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from an external API, creating a potential surface for indirect prompt injection. 1. Ingestion points: External search results and AI-generated answers are fetched from https://openapi.felo.ai/v2/chat via the search.sh script. 2. Boundary markers: The skill instructs the agent to display the output under specific markdown headers, but it lacks explicit delimiters or instructions to ignore potential commands embedded in the fetched content. 3. Capability inventory: The skill utilizes the bash tool to perform network operations and local script execution. 4. Sanitization: There is no evidence of sanitization or filtering of the API's response before it is integrated into the agent's context.
[EXTERNAL_DOWNLOADS]: The documentation suggests installing the skill using the agent platform's command-line tool which fetches code from the official vendor repository on GitHub. This fetches configuration and code from the official vendor infrastructure.

felo-search