platform-custom-metadata-type-generate
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external sources to generate Salesforce metadata, creating a surface for indirect prompt injection attacks.
- Ingestion points: SKILL.md (Section 4) describes workflows where the agent reads data from inline tables, CSV files, or prose descriptions provided by the user.
- Boundary markers: No specific boundary markers or instructions to ignore embedded commands within the ingested data are mandated in the instructions.
- Capability inventory: The agent has access to the Salesforce CLI (
sf) and the ability to write multiple.md-meta.xmlfiles to the local filesystem. - Sanitization: The skill mitigates part of the risk by using a local utility script,
scripts/sanitize-developer-name.sh, to ensure that record names conform to strict character and length constraints before generation. - [COMMAND_EXECUTION]: The skill utilizes the Salesforce CLI (
sf) as a primary tool for metadata operations. This is a standard and expected capability for the skill's purpose as an administrative tool for the Salesforce platform. - [DYNAMIC_EXECUTION]: The skill instructs the agent to execute a locally provided bash script (
scripts/sanitize-developer-name.sh) to process and sanitize user input. This script performs deterministic string transformations (usingsedandcut) to generate valid Salesforce DeveloperNames, which is a low-risk use of dynamic execution for data normalization.
Audit Metadata