All 42 static findings are false positives. They match Markdown formatting, Azure placeholders, or documented environment configuration. No malicious intent or prompt injection appears.
The static alerts are false positives caused by markdown code fences, inline code formatting, placeholder URLs, and normal Azure SDK configuration examples. No prompt injection, hidden command execution, credential exfiltration, or system reconnaissance intent was found in SKILL.md.
Static analysis flagged Markdown fences, inline backticks, placeholder URLs, and Azure environment variable examples. All static findings are false positives because the skill is documentation for Azure AI Vision SDK usage, with no executable files or malicious intent. No evidence found of prompt injection, credential exfiltration, or system reconnaissance.
The static analyzer heavily over-reported Markdown code fences as Ruby or shell backtick execution, and the weak-crypto and reconnaissance hits are false positives. The skill is still medium risk because it documents installing an external package, reading Azure credentials from environment variables, and sending image data or image URLs to Azure AI Vision.
1
Files scanned
261
Lines analyzed
7
Review items
3
False positives ignored
Capability review items (4)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
The examples configure an Azure Cognitive Services endpoint and call Azure AI Vision analysis methods for URLs and local image bytes. This is expected behavior for the skill, but users may send sensitive images or image URLs to a cloud service if they run the examples without reviewing data handling requirements.
The network behavior is explicit in the Azure endpoint and analyze_from_url/analyze examples. It appears legitimate for an Azure Vision SDK guide, with no evidence of hidden or unauthorized destinations.
The examples read VISION_ENDPOINT and VISION_KEY from environment variables and also show DefaultAzureCredential. This is a standard Azure pattern, but it requires users to protect API keys and avoid exposing credentials in logs, prompts, or shared files.
The credential reads are direct and visible in Python examples. The usage is normal for Azure SDK authentication, but credentials are security-sensitive by nature.
The skill instructs users to install azure-ai-vision-imageanalysis with pip. This is a normal setup step, but it introduces supply-chain and dependency trust considerations for users who run the command.
The pip install command is clear and limited to the declared Azure Vision package. The risk is dependency-related rather than evidence of malicious command execution.
The critical heuristic combined code execution, network, and credential access signals. Review found visible SDK examples for Azure authentication and image analysis, but no obfuscation, prompt injection, credential exfiltration, or hidden malicious workflow.
The risk factors exist in documentation examples, but the context supports the advertised Azure Vision purpose. No suspicious instructions, disguised endpoints, or prompt-injection text were found.
Static false positives ignored (3)
These static matches were dismissed by semantic review or matched schema-only tokens, so they are shown for transparency but do not drive the quality score.
Markdown Backtick Execution Findings Are False Positives
The static Ruby or shell backtick findings point to Markdown code fences, inline code, and documentation examples. No evidence found of Ruby backtick execution or hidden script execution in SKILL.md.
Line review shows Markdown fences and Python or bash examples, not executable Ruby backtick syntax. The only actual shell instruction is the visible pip install example.
The weak cryptographic algorithm findings do not correspond to cryptographic code, hashing, encryption, or certificate handling. No evidence found of MD5, SHA-1, DES, RC4, or similar weak algorithms in the reviewed file.
The cited lines contain metadata, a Markdown table, and a best-practice sentence. They do not contain cryptographic operations or algorithm names.
The system reconnaissance finding points to a best-practice line about using async clients for high-throughput scenarios. No evidence found of host probing, OS enumeration, environment dumping, or system discovery commands.
The cited line is plain documentation text about performance. It does not request or execute any system reconnaissance activity.
This is a prompt-only documentation skill containing reference material for the Azure AI Vision Image Analysis SDK. No executable code was detected. The skill provides documentation on how to use Microsoft's Azure computer vision service for image analysis tasks including captioning, object detection, OCR, and smart cropping. No security concerns identified.
0
Files scanned
0
Lines analyzed
0
Review items
0
False positives ignored
No confirmed security findings were recorded for this completed audit.