The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute complex Bash commands, including loops and pipelines involving curl and jq, to fetch procurement data from external sources and local files.
[INDIRECT_PROMPT_INJECTION]: The skill processes and displays data retrieved from external government APIs (such as project descriptions and financial details) without explicit boundary markers or sanitization, creating a potential vector for instructions embedded in the source data to influence agent behavior.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to multiple Italian government data portals, including api.sogei.it, bdap-opendata.rgs.mef.gov.it, www.serviziocontrattipubblici.it, and opencoesione.gov.it, to retrieve project and tender information.
[COMMAND_EXECUTION]: The instructions explicitly direct the use of the -k (or --insecure) flag with curl when accessing www.serviziocontrattipubblici.it due to a reported self-signed certificate. Bypassing SSL verification exposes the communication to potential Man-in-the-Middle (MitM) attacks.

cup-cig