The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is designed to ingest and process external project data from various formats (CSV, Excel, JSON). There are no explicit sanitization or boundary markers mentioned to prevent the agent from interpreting instructions embedded within this external data.
Ingestion points: Processes user-provided project data in generate_progress_report, generate_cost_report, and generate_safety_report methods within SKILL.md.
Boundary markers: None detected; the skill relies on standard data structures without explicit 'ignore' instructions for the content of data fields.
Capability inventory: The skill possesses filesystem write capabilities via export_to_excel and generates HTML strings via to_html.
Sanitization: No evidence of data sanitization or instruction filtering for incoming data streams.
[COMMAND_EXECUTION]: The export_to_excel method in SKILL.md utilizes pandas.ExcelWriter to write data to a file path provided as an argument. While intended for report generation, this presents a risk of arbitrary file write if the agent is manipulated into using a sensitive system path as the output destination.
[DATA_EXPOSURE]: The to_html method in SKILL.md generates an HTML report by directly interpolating data values into HTML strings (e.g., f"<td>{value}</td>"). This lacks proper HTML escaping, creating a vulnerability to HTML injection or Cross-Site Scripting (XSS) if the generated report is rendered in a web browser or HTML viewer.

pdf-report-generator