The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The generate_html_gantt method in SKILL.md interpolates task data directly into HTML templates without sanitization.
Ingestion points: Task data is read from user-provided files (Excel/CSV) via the import_from_df method.
Boundary markers: The skill lacks explicit instructions to prevent the agent from executing instructions hidden in the task data.
Capability inventory: The skill possesses file system write capabilities through the export_to_excel and generate_html_gantt functions.
Sanitization: No HTML escaping is applied to user-controlled strings before report generation.
[COMMAND_EXECUTION]: The skill implements export_to_excel in SKILL.md to write files to the local system, utilizing the filesystem permission defined in claw.json.

gantt-chart