The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill processes data from exploration threads and charts to build narratives. This is the primary function of the skill and is handled through structured data inspection tools rather than direct interpolation of untrusted strings into executable contexts.
[Token Leakage Prevention]: The code implementation includes a proactive cleaning function (_strip_leaked_tool_syntax) that uses regular expressions to remove internal model tokens or leaked tool call syntax from the final report. This provides a layer of defense-in-depth to maintain output quality and security.
[Structured Data Inspection]: The inspect_chart tool leverages the pandas library to provide the agent with structured summaries and samples of data. This allows the agent to reason about the data without requiring direct access to underlying raw files or executing arbitrary code on the datasets.

report