platform-custom-field-generate

Pass

Audited by Gen Agent Trust Hub on Nov 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest existing project metadata and SObject schemas to inform the generation of new field definitions, creating a potential attack surface for indirect instructions.
  • Ingestion points: Section 6 of SKILL.md requires the agent to read all local **/*.field-meta.xml files and retrieve SObject schemas.
  • Boundary markers: The instructions do not mandate the use of delimiters or specific directives to ignore instructions potentially embedded in the ingested data.
  • Capability inventory: The skill allows for the creation of XML metadata files and utilizes the search_metadata tool.
  • Sanitization: The skill requires wrapping formulas in CDATA blocks within the XML to ensure special characters are handled as data rather than markup.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 7, 2026, 11:16 AM