The Agent Skills Directory

[SAFE]: The skill implements a reporting mechanism that uses the SendMessage tool to communicate status updates to a recipient defined in the agent's mandate. The instructions focus on structured data integrity and efficient context management.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from a 'mandate' and incorporates 'established facts' from previously executed commands into its reports. While this creates a theoretical surface for indirect prompt injection if those sources contain malicious instructions, the skill enforces a rigid schema for the output packet which acts as a structural boundary. The ingestion points include the mandate fields and tool outputs, and while explicit sanitization is not defined, the strict 'one fact per line' and 'command + fact' pairing constraints limit the potential for control-flow hijacking.

report