The Agent Skills Directory

[PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection via the processing of external data files into an HTML dashboard.
Ingestion points: Untrusted data enters the agent context through data/transitions.csv and data/txns.csv, which are used to populate model results.
Boundary markers: Absent. There are no explicit instructions or delimiters used to prevent the agent from interpreting embedded instructions or malicious strings within the CSV content.
Capability inventory: The skill utilizes bash_tool for running Python calculations and show_widget for rendering the references/ltv_dashboard_template.html dashboard.
Sanitization: The HTML template ltv_dashboard_template.html uses innerHTML to inject state labels and KPI strings directly into the DOM (e.g., in the KPI bar and the period detail table). If the source CSV contains malicious HTML/JavaScript (e.g., <img src=x onerror=alert(1)>), it will execute within the dashboard widget's context.
[COMMAND_EXECUTION]: The skill's workflow involves executing multiple local Python scripts via bash_tool (e.g., ltv_descriptive.py, ltv_markov_chain.py, ltv_survival.py) to perform complex data analysis. This is documented and aligned with the skill's primary purpose of providing modeling capabilities.

customer-lifetime-value