The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external ledger data through the scripts/import_csv.mjs script, which is then rendered in the AirApp dashboard. This creates a surface where malicious instructions embedded in a CSV file could potentially influence the agent's behavior when analyzing the fund.
Ingestion points: Data enters the system via the scripts/import_csv.mjs command-line utility which parses user-provided CSV files.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard natural language commands found within the imported ledger data.
Capability inventory: The skill can read from and write to Busabase Bases and perform financial fairness calculations. It is explicitly designed to be read-only in the UI and lacks capabilities for money transfers or arbitrary shell execution.
Sanitization: The frontend application (app.js) uses an escapeHtml function to sanitize data before it is rendered in the user interface, mitigating cross-site scripting risks.
[COMMAND_EXECUTION]: The skill provides utility scripts such as scripts/import_csv.mjs, scripts/setup.mjs, and scripts/publish_airapp.mjs. These are intended for the user to execute manually in their environment to manage the skill's data and deployment. They rely on environment variables for credentials, which is a standard and secure practice.

kelly-family-fund