The Agent Skills Directory

[Indirect Prompt Injection]: The skill is designed to ingest and process external financial data such as trial balances and subledger details. This creates a surface for indirect prompt injection if the source data provided by the user contains malicious instructions. The inclusion of a manual review checklist in the skill serves as an important human-in-the-loop mitigation.
[No Executable Code]: Analysis of the skill reveals no scripts, binary files, or command-line execution patterns. The skill functions as a template for data transformation within the agent's conversational context.

journal-entry