kelly-restaurant-intel
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data to generate business signals, operating actions, and marketing drafts, creating a surface for indirect prompt injection attacks.
- Ingestion points: The skill ingests data from public news sources, local events, trend signals, competitor menus, and customer reviews (referenced in
SKILL.mdandREADME.md). - Boundary markers: The
SKILL.mdfile defines explicit boundaries, instructing the agent to block allergen/food-safety claims unless sourced and to require human approval for all customer-visible drafts or regulated claims. - Capability inventory: The skill can write signals and drafts to Busabase, run a local Hono-based web server (
server.js), and execute decision-tracking scripts (execute_decisions.mjs). - Sanitization: The web application implements HTML escaping for rendered content in
app.jsandaccent-theme.jsto prevent cross-site scripting (XSS) when displaying ingested data.
Audit Metadata