The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data such as news, weather, events, and competitor menus to generate business insights and marketing drafts. This creates an indirect prompt injection surface where malicious content in those external sources could attempt to influence the agent's output. The skill mitigates this through explicit safety instructions and by requiring human review and approval for all actions and drafts before any external handoff occurs.
Ingestion points: The agent collects data from public URLs, news sources, and competitor websites as defined in the SKILL.md workflow.
Boundary markers: The skill instructions explicitly mandate blocking allergen and food-safety claims unless sourced, providing a logical boundary for the agent's reasoning.
Capability inventory: The skill uses the busabase-sdk for database operations and includes a script scripts/execute_decisions.mjs that allows the user to mark approved items as done. It does not perform autonomous external actions like posting to social media or sending emails.
Sanitization: The web interface implements escapeHtml in app/app/app.js to sanitize data retrieved from the database before rendering it in the user's browser.
[SAFE]: The skill uses the Busabase SDK for secure data management within a defined workspace. Credentials and sensitive operations are handled according to standard platform patterns, and the agent's capabilities are scoped to reading and writing to its own database bases within the Busabase environment.

kelly-restaurant-intel