inventory-planning

Pass

Audited by Gen Agent Trust Hub on Sep 25, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it is designed to ingest data from untrusted external sources and incorporate that data into its persistent local context.
  • Ingestion points: The skill processes external SKU-level inventory exports, sales history, promotional calendars, and content fetched from public websites during company research. This untrusted data is directly utilized to generate analysis and update the agent's context.
  • Boundary markers: The instructions do not explicitly require the agent to use strict delimiters (like XML tags) or specific boundary markers to isolate untrusted data from its primary instructions, increasing the risk that embedded commands in the data could be interpreted as agent directives.
  • Capability inventory: The skill has authorized capabilities to read and write to the local filesystem (specifically .agents/product-context.md), perform network operations to fetch external URLs, and invoke platform-provided tools such as MCP servers and UI Artifacts.
  • Sanitization: There are no explicit requirements for sanitizing, filtering, or escaping external content before it is processed or written to the product-context file, though the skill does reference general 'Safety' guidelines in its house-rules for treating fetched content as data rather than instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 25, 2026, 04:06 PM