The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and re-pitch the "last message" in a conversation, which constitutes an ingestion point for untrusted data.
Ingestion points: The agent's previous conversation history (referred to as "that last message") is the primary input.
Boundary markers: The instructions do not use specific delimiters or protective framing to separate the re-pitching instructions from the content of the message being processed.
Capability inventory: The skill has no active capabilities; it does not request tools (allowed-tools is empty/absent), execute scripts, or perform network operations. The disable-model-invocation: true flag further limits the skill to manual triggers.
Sanitization: There is no evidence of input filtering or sanitization, though the instruction to use ASD-STE100 Simplified Technical English naturally constrains the output format.

wait-what