load-standards
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is configured to read 22 markdown files from the local
standards/directory to provide context for connector development. This data ingestion creates an indirect prompt injection surface where instructions potentially embedded within those files could attempt to override agent behavior.\n - Ingestion points: The
SKILL.mdworkflow reads 22 markdown files (e.g.,standards/main.md,standards/source_types/sql_databases.md) into the active context.\n - Boundary markers: The skill includes a 'Content Trust' section in
SKILL.mdthat explicitly instructs the agent to ignore any instructions found within the loaded files and treat the content as reference data only.\n - Capability inventory: The skill's primary capabilities involve reading local files and generating descriptive summaries of their content.\n
- Sanitization: No automated content validation, filtering, or schema enforcement is applied to the ingested markdown files before they are processed by the agent.
Audit Metadata