migrate-to-dataverse

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • Indirect Prompt Injection Surface: The skill processes .pa.yaml files, which are the source code files of a Canvas App. These files act as external data that the agent must parse and transform, creating a potential surface for indirect instructions embedded within the app's formulas or metadata.\n
  • Ingestion points: The agent reads all local .pa.yaml files generated by the sync_canvas tool to identify Power Fx data source calls.\n
  • Boundary markers: The skill does not define specific delimiters to separate the YAML content from the agent's instructions, though it targets specific code patterns like Filter or Patch calls.\n
  • Capability inventory: The skill possesses the ability to read and write files, and it can execute platform-specific tools for syncing, schema discovery, and compilation.\n
  • Sanitization: The skill implements a security best practice by requiring a Proposed Mapping Plan to be reviewed and approved by the user before any file modifications are executed. It also includes a validation step using compile_canvas to ensure the integrity of the updated app.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 07:03 AM
Security Audit — agent-trust-hub — migrate-to-dataverse