The Agent Skills Directory

[SAFE]: The skill documents standard usage of the iii engine primitives for reliable asynchronous task execution. No malicious code, obfuscation, or exfiltration patterns were detected.\n- [INDIRECT_PROMPT_INJECTION]: The skill describes a pattern where background workers process job payloads. This creates a surface for indirect prompt injection if the data being processed originates from untrusted sources.\n
Ingestion points: Payloads received by registerFunction handlers (SKILL.md).\n
Boundary markers: Absent; no specific instructions provided for delimiting untrusted content.\n
Capability inventory: Uses trigger for enqueuing and registerFunction for task processing (SKILL.md).\n
Sanitization: Not addressed in the architectural pattern description.

queue-processing