The Agent Skills Directory

Unverifiable Dependencies & Remote Code Execution (LOW): The skill utilizes the Maven dependency com.azure:azure-ai-vision-imageanalysis version 1.1.0-beta.1.
Evidence: References found in SKILL.md and acceptance-criteria.md.
Notes: Microsoft/Azure is a trusted organization; the finding is downgraded per the [TRUST-SCOPE-RULE].
Indirect Prompt Injection (LOW): The skill processes untrusted image content which could contain malicious text or instructions intended to influence the agent.
Ingestion points: The analyzeFromUrl and analyze methods in SKILL.md accept arbitrary input from external URLs and local files.
Boundary markers: No delimiters or instruction-ignore guards are used around the extracted text.
Capability inventory: Capability is currently limited to terminal display (System.out.printf) of extracted text.
Sanitization: No sanitization or validation of the OCR output or captioning is performed.
Data Exposure & Exfiltration (LOW): The skill performs network operations to Azure Cognitive Services endpoints.
Evidence: Network usage is central to the ImageAnalysisClient implementation in SKILL.md.
Notes: This is a functional requirement; no access to or exfiltration of sensitive system files was detected.

azure-ai-vision-imageanalysis-java