skills/datex/skills/package-cascade/Gen Agent Trust Hub

package-cascade

Pass

Audited by Gen Agent Trust Hub on Aug 31, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data from the Datex marketplace and user-provided inputs to automate repository updates and package publishing.
  • Ingestion points: User-supplied package names and versions (SKILL.md), and JSON output from dxs marketplace search and dxs source cascade plan (references/cascade-workflow.md).
  • Boundary markers: The skill requires resolving display names to uniqueIdentifiers using tool-based search before planning and mandates explicit user verification prompts for all plan execution and identity switching actions.
  • Capability inventory: Performs branch creation, configuration re-pinning, commits, and package publishing via dxs source cascade run. It also manages authentication flows using dxs auth login.
  • Sanitization: The instructions strictly delegate all data handling and configuration modification to the dxs CLI tool, prohibiting the agent from manually editing AppConfig JSON files to prevent data corruption or injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 31, 2026, 06:14 PM
Security Audit — agent-trust-hub — package-cascade