sbom-analysis
Pass
Audited by Gen Agent Trust Hub on Jul 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes external, potentially untrusted SBOM and VEX data which could contain embedded instructions intended to influence the agent's behavior (Indirect Prompt Injection).
- Ingestion points: SBOM files (CycloneDX, SPDX) and VEX documents (CSAF, OpenVEX) identified in the 'Context the Agent Needs' section and 'Step 1' of the process.
- Boundary markers: The skill does not define structural delimiters for the untrusted content, but it does provide a 'Prompt Injection Safety Notice' directing the agent to ignore instructions embedded in metadata or justifications.
- Capability inventory: The skill is restricted to using 'Read', 'Grep', and 'Glob' tools as defined in the frontmatter. It lacks network access or file-writing capabilities.
- Sanitization: No programmatic sanitization is defined; the skill relies on the agent's adherence to the safety notice to disregard commands within the input data.
Audit Metadata