Security audit
Security checks for vulnerabilities and agentic risk
This skill generates draw.io diagram files and exports images using local draw.io commands; its behavior is coherent with that purpose, with some usability and consent caveats.
Install only if you want the agent to create local diagram artifacts and run the draw.io exporter. Ask the agent to confirm filenames and formats first if you do not want automatic files under ./diagrams, and be aware the skill may activate from broad diagram-related terms.
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
<mxCell id="0"/>
<mxCell id="1" parent="0"/>
<!-- CLASS: Address -->
<mxCell id="10" value="Address"
style="swimlane;fontStyle=1;align=center;startSize=30;fontSize=14;fillColor=#dae8fc;strokeColor=#6c8ebf;"
vertex="1" parent="1">
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
vertex="1" parent="10">
<mxGeometry y="130" width="220" height="10" as="geometry"/>
</mxCell>
<!-- Methods -->
<mxCell id="13" value="-validate()
+outputAsLabel()"
style="text;strokeColor=none;fillColor=none;align=left;verticalAlign=top;spacingLeft=4;spacingRight=4;overflow=hidden;rotatable=0;fontSize=12;"
vertex="1" parent="10">
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
<mxGeometry y="60" width="220" height="10" as="geometry"/>
</mxCell>
<!-- RELATIONSHIP: Address -> Person (association "lives at" 0..1) -->
<mxCell id="80" value="lives at"
style="edgeStyle=orthogonalEdgeStyle;endArrow=open;endFill=0;startArrow=none;fontSize=11;"
edge="1" parent="1" source="10" target="20">
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
<mxGeometry relative="1" as="geometry"/>
</mxCell>
<!-- RELATIONSHIP: Professor extends Person (inheritance - hollow triangle) -->
<mxCell id="83" value=""
style="edgeStyle=orthogonalEdgeStyle;endArrow=block;endFill=0;startArrow=none;fontSize=11;"
edge="1" parent="1" source="40" target="20">
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
<mxCell id="0"/>
<mxCell id="1" parent="0"/>
<!-- TABLE: Cliente | height = 30 + (5 x 30) = 180 -->
<mxCell id="10" value="Cliente"
style="shape=table;startSize=30;container=1;collapsible=0;childLayout=tableLayout;fixedRows=1;rowLines=0;fontStyle=1;fontSize=14;align=center;fillColor=#dae8fc;strokeColor=#6c8ebf;fontColor=#000000;swimlaneLine=1;"
vertex="1" parent="1">
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
<mxGeometry x="40" width="180" height="30" as="geometry"><mxRectangle width="180" height="30" as="alternateBounds"/></mxGeometry>
</mxCell>
<!-- Row: Apellidos -->
<mxCell id="17" value=""
style="shape=tableRow;horizontal=0;startSize=0;swimlaneHead=0;swimlaneBody=0;fillColor=none;collapsible=0;dropTarget=0;points=[[0,0.5],[1,0.5]];portConstraint=eastwest;fontSize=12;top=0;left=0;right=0;bottom=0;"
vertex="1" parent="10">
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
<mxGeometry x="40" width="180" height="30" as="geometry"><mxRectangle width="180" height="30" as="alternateBounds"/></mxGeometry>
</mxCell>
<!-- Row: EMail -->
<mxCell id="23" value=""
style="shape=tableRow;horizontal=0;startSize=0;swimlaneHead=0;swimlaneBody=0;fillColor=none;collapsible=0;dropTarget=0;points=[[0,0.5],[1,0.5]];portConstraint=eastwest;fontSize=12;top=0;left=0;right=0;bottom=0;"
vertex="1" parent="10">
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
<mxGeometry x="80" y="380" width="220" height="120" as="geometry"/>
</mxCell>
<!-- Row: ID (PK) -->
<mxCell id="31" value=""
style="shape=tableRow;horizontal=0;startSize=0;swimlaneHead=0;swimlaneBody=0;fillColor=#e1d5e7;collapsible=0;dropTarget=0;points=[[0,0.5],[1,0.5]];portConstraint=eastwest;fontSize=12;top=0;left=0;right=0;bottom=1;"
vertex="1" parent="30">
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
<mxGeometry x="40" width="180" height="30" as="geometry"><mxRectangle width="180" height="30" as="alternateBounds"/></mxGeometry>
</mxCell>
<!-- Row: ClienteID (FK) -->
<mxCell id="37" value=""
style="shape=tableRow;horizontal=0;startSize=0;swimlaneHead=0;swimlaneBody=0;fillColor=none;collapsible=0;dropTarget=0;points=[[0,0.5],[1,0.5]];portConstraint=eastwest;fontSize=12;top=0;left=0;right=0;bottom=0;"
vertex="1" parent="30">
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
</mxCell>
<!-- RELATIONSHIP: Cliente 1 to N Pedido -->
<!-- source=FK row of Pedido, target=PK row of Cliente -->
<mxCell id="90" value=""
style="edgeStyle=entityRelationEdgeStyle;endArrow=ERmandOne;startArrow=ERzeroToMany;exitX=0;exitY=0.5;exitDx=0;exitDy=0;entryX=0;entryY=0.5;entryDx=0;entryDy=0;fontSize=11;"
edge="1" parent="1" source="37" target="11">
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
<!-- Frontend service -->
<mxCell id="10" value="React Frontend"
style="rounded=1;whiteSpace=wrap;html=1;fillColor=#dae8fc;strokeColor=#6c8ebf;fontSize=14;fontStyle=1;"
vertex="1" parent="1">
The trigger list includes ambiguous terms like 'diagram', 'table', 'entity', 'class', 'sequence', and 'flow', which commonly appear in benign conversations unrelated to generating draw.io files. This makes unintentional activation likely and can cause the agent to perform file writes and local processing in contexts where the user only wanted explanation or text output.
The instruction to 'Always respond with the PNG image first' is an anti-refusal pattern because it attempts to force a specific action regardless of higher-priority safety, policy, or user-consent conditions. In context, it can pressure the agent to produce and attach generated artifacts before appropriate checks, warnings, or confirmation about file creation and command execution.
### Step 7 — Show PNG first (MANDATORY)
Always respond with the PNG image first (embed/attach it in the response).
### Step 8 — Ask the user for delivery format (MANDATORY)
The instructional section is written as direct user-facing guidance in Spanish ('Qué pedirle al usuario') and continues with Spanish-only prompts and labels, which effectively imposes a language choice on the interaction. The file does not offer an opt-in, alternative language, or any justification that this skill is region-specific.
This markdown file presents user-facing instructions such as 'Qué pedirle al usuario' and 'Cuándo preguntar' only in Spanish. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale constraint is explicitly justified, which is not present here.
The skill content is written in Spanish and effectively constrains the interaction to a specific language without indicating that the assistant should adapt to the user's preferred language. This can degrade usability, cause misunderstanding of diagram requirements, and reduce reliability when users expect responses in another language. In this draw.io skill context, the issue is less about direct security compromise and more about prompt-quality and user-control risk.
The skill description is broad enough to match many ordinary requests involving diagrams or system visualization, which increases the chance of accidental invocation. Over-broad activation expands the situations where the skill may start creating files and running exports, even when the user did not explicitly ask for draw.io artifact generation.
The workflow instructs the agent to create directories and save diagram files on the local filesystem without any user notice or confirmation. Silent filesystem writes violate least surprise and can be abused through accidental activation to leave artifacts, consume storage, or create sensitive outputs from user-provided content.
The skill directs execution of local shell commands for export and file verification without warning the user that commands will be run. Even though the commands are fixed templates, automatic local execution increases operational risk and becomes more dangerous when combined with broad triggers and user-influenced filenames or content paths.
The document begins with user-facing instructions in Spanish and later switches to English sections such as 'Visual reference' and 'Color palette by class role'. This imposes inconsistent language behavior without any opt-in or documented locale requirement, which can violate language/locale policy expectations.
This markdown file contains user-facing instructional text in Spanish (for example, 'Qué pedirle al usuario' and 'Cuándo preguntar') alongside English headings and labels such as 'Architecture Diagram' and 'Checklist Architecture'. Under the policy rule, forcing or assuming a language/locale without opt-in can be a natural-language policy issue when no language choice or justification is provided.
No suspicious patterns detected.