data-integrity-failures
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and analyze content from untrusted external sources including user uploads, external APIs, and message queues which may contain embedded instructions.
- Ingestion points: Data categorized as untrusted sources in the Software and Data Integrity Failures Skill section and when_to_use metadata field.
- Boundary markers: The skill does not define explicit delimiters for the agent to use when processing these external data sources.
- Capability inventory: The skill is intended for an agent with capabilities to write and modify source code, manage CI/CD pipelines, and handle package installations.
- Sanitization: The skill mandates strict sanitization by requiring the use of schema-validated formats such as JSON or protobuf and forbidding unsafe deserialization methods like Python's pickle or Ruby's Marshal.
Audit Metadata