Video Call Agent

• Outcome: Pass
• Latest audit: May 28, 2026 · 12:24 PM UTC
• Version: 1.0.0

Overview

This skill is a disclosed Runway-powered video call workflow, but users should treat calls, transcripts, recordings, avatar images, and personality prompts as sensitive data shared with external services.

Install only if you are comfortable using Runway and the disclosed npm runtime for live audio/video calls. Avoid putting secrets, unnecessary personal details, private project data, or biometric face images into avatar prompts or uploads unless needed; use cloudflared tunnels only when required; ask before sharing recordings or detailed summaries in other channels; and stop the local server and delete unused avatars when finished.

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (7)

Vague Triggers

Medium

• Confidence: 85% confidence
• Finding: The skill is marked user-invocable and describes broadly starting a video call for many ordinary situations, which can cause over-triggering for routine interactions that would be safer and less intrusive as text. Because the action initiates real-time audio/video collection and can lead to transcript/recording generation, an overly broad invocation surface materially increases privacy and consent risk.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The skill instructs the agent to send call recordings and summarize call contents without placing a prominent upfront warning and consent requirement before those steps. That creates a real risk that sensitive spoken content will be recorded, retained, and redistributed to a chat channel by default, even when the user may not expect or want that handling.

Ssd 3

Medium

• Confidence: 95% confidence
• Finding: The instructions encourage embedding remembered user context, preferences, projects, and other personal details into the avatar's persistent personality prompt. Storing or repeatedly reusing that context in a long-lived avatar increases unnecessary retention and risks disclosure of personal or sensitive information across future calls or to the external provider.

Ssd 3

Medium

• Confidence: 96% confidence
• Finding: The skill directs the agent to download the recording and send it back through the user channel by default, along with a summary derived from the transcript. This broadens disclosure beyond the live call itself and can expose sensitive spoken information in channels that may have different retention, visibility, or access controls than the call.

External Transmission

Medium

• Category: Data Exfiltration

Content

4. **Generate one with Runway text-to-image** — if none of the quick picks fit your character, or if the user specifically asks for a custom avatar look. Use the Runway API with model `gemini_2.5_flash`:

```bash
curl -X POST https://api.dev.runwayml.com/v1/text_to_image \
  -H "Authorization: Bearer $RUNWAYML_API_SECRET" \
  -H "X-Runway-Version: 2024-11-06" \
  -H "Content-Type: application/json" \

• Confidence: 90% confidence
• Finding: The referenced Runway API endpoint is an explicit third-party destination, so the skill does involve external transmission. In the context of this skill, that matters because prompts may include personality and user context, making unintended disclosure more likely if the agent over-shares.

External Transmission

Medium

• Category: Data Exfiltration

Content

4. **Generate one with Runway text-to-image** — if none of the quick picks fit your character, or if the user specifically asks for a custom avatar look. Use the Runway API with model `gemini_2.5_flash`:

```bash
curl -X POST https://api.dev.runwayml.com/v1/text_to_image \
  -H "Authorization: Bearer $RUNWAYML_API_SECRET" \
  -H "X-Runway-Version: 2024-11-06" \
  -H "Content-Type: application/json" \

• Confidence: 90% confidence
• Finding: The referenced Runway API endpoint is an explicit third-party destination, so the skill does involve external transmission. In the context of this skill, that matters because prompts may include personality and user context, making unintended disclosure more likely if the agent over-shares.

External Transmission

Medium

• Category: Data Exfiltration

Content

**Step 1: Create an upload URL**

```bash
curl -X POST https://api.dev.runwayml.com/v1/uploads \
  -H "Authorization: Bearer $RUNWAYML_API_SECRET" \
  -H "X-Runway-Version: 2024-11-06" \
  -H "Content-Type: application/json" \

• Confidence: 90% confidence
• Finding: The upload flow sends image data to an external provider, which is a real data transmission risk because user-supplied images may contain biometric or personal information. This is especially sensitive in an avatar-creation workflow where face images are likely and may persist or be processed externally.

Static analysis

No suspicious patterns detected.