The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a vulnerability surface for indirect prompt injection by incorporating processed data into a Markdown report without boundary markers or complete sanitization.
Ingestion points: Untrusted data enters the agent context through the df object, which is derived from external files and processed in SKILL.md.
Boundary markers: Absent. The report generation logic in Step 5 appends values from the data directly into the Markdown string without using delimiters or instructions for the agent to ignore embedded content.
Capability inventory: The skill performs file system writes (to_excel in Step 3, savefig in Step 4) and generates structured analysis reports across all steps.
Sanitization: Partial. While Step 1 uses regex to clean the target_col, other fields such as group_col and the data exported via to_markdown() in Step 5 are interpolated into the prompt without validation or escaping.

chart-embedded-export