Veevid Skill

• Outcome: Review
• Latest audit: Oct 15, 2026 · 3:36 PM UTC
• Version: 1.3.1

Overview

This is a disclosed Veevid API helper, but it can send sensitive media and identity-manipulation inputs to an external service without enough consent or privacy guardrails.

Review this skill carefully before installing. It is not showing hidden code or malware behavior, but using it means sending prompts, API keys, and selected media or documents to Veevid. Only use photos, voices, recordings, and documents you are authorized to upload, avoid non-consensual or deceptive real-person edits, and protect the API key stored in ~/.config/veevid/api_key.

A.I.G

Vulnerability Patterns

• Skill Instruction Hijacking: Alters the agent's session goals or safety constraints when the skill loads
• Agent Memory Poisoning: Writes attacker-controlled rules into memory that affect later sessions
• Remote Payload Retrieval and Execution: Fetches external code whose behavior can change after review
• Embedded Malicious Code: Ships malicious scripts inside the skill and executes them locally
• Unauthorized Access and Privilege Escalation: Obtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns

• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Rogue Agent: Self-Modification, Session Persistence
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection

Findings (15)

Missing User Warnings

High

• Category: Not specified by scanner
• Confidence: 95% confidence

Finding

Lip-sync mode processes a portrait image together with speech audio, which can involve biometric facial data and voice data, yet the documentation gives no warning about sending this sensitive data to a third-party service. That omission increases the risk of privacy violations, non-consensual deepfake creation, and regulatory exposure when users submit their own or others' likenesses and voices without informed consent.

Content

No source excerpt is available for this finding.

Session Persistence

Medium

• Category: Rogue Agent
• Confidence: 60% confidence

Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 44)May include surrounding context.

2. Save it:

bash

mkdir -p ~/.config/veevid
echo "vv_sk_your_key" > ~/.config/veevid/api_key

External Transmission

Medium

• Category: Data Exfiltration
• Confidence: 60% confidence

Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 61)May include surrounding context.

bash

KEY=$(cat ~/.config/veevid/api_key)
BODY='{"mode":"veo3","generation_type":"text-to-video","prompt":"A golden retriever running through sunflowers, cinematic","video_quality":"standard","aspect_ratio":"16:9"}'
curl -X POST https://veevid.ai/api/quote -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" -d "$BODY"
# → {"required_credits":<n>,"current_balance":<n>,"sufficient":true}
# STOP: tell the user the model and <n> credits, and only continue after they say yes.

Natural-Language Policy Violations

Medium

• Category: Not specified by scanner
• Confidence: 95% confidence

Finding

The documentation says 'Write prompts in English (translate the user's request yourself),' which imposes a specific language choice. This is a natural-language locale policy concern because it does not offer the user a language choice or explain a justified region-specific requirement.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

• Category: Not specified by scanner
• Confidence: 95% confidence

Finding

The text explicitly instructs users to 'write in English,' which imposes a language requirement. This is a natural-language policy concern because it forces a specific language without offering a choice or documenting a justified locale limitation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium

• Category: Not specified by scanner
• Confidence: 91% confidence

Finding

The speech-to-text documentation describes uploading audio to an external service and returning a transcript, but it does not clearly warn that source audio URLs are retained for 14 days and that transcript content may contain sensitive personal or confidential data. This omission can cause users or downstream agents to submit regulated or sensitive audio without informed consent or appropriate safeguards, creating privacy and compliance risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium

• Category: Not specified by scanner
• Confidence: 88% confidence

Finding

The audio cleanup section instructs users to send audio or video files to an external processing service but does not prominently warn that sensitive spoken content, background sounds, or embedded metadata may be exposed to third-party processing. Users may assume this is a local or low-risk transformation and inadvertently transmit confidential recordings.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium

• Category: Not specified by scanner
• Confidence: 91% confidence

Finding

This documentation explicitly supports uploading a person's photo and garment images for virtual try-on, but it provides no privacy, consent, retention, or sensitive-biometric handling guidance. Because the skill is specifically designed to process identifiable personal images, omission of consent and privacy safeguards can lead to unauthorized use of someone's likeness and privacy-law or policy violations.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium

• Category: Not specified by scanner
• Confidence: 91% confidence

Finding

This section similarly instructs users to upload a person's photo for product-holding generation without any warning about consent, privacy, or authorized use of the image. In the context of an image-generation API, this makes misuse easier by normalizing processing of identifiable photos without safeguards or disclosure expectations.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium

• Category: Not specified by scanner
• Confidence: 91% confidence

Finding

This section documents an AI kissing video generator for intimate content but provides no safety guidance about consent, age verification, or misuse such as creating non-consensual sexualized deepfakes. In a skill specifically designed to help agents invoke these endpoints, omission of such guardrails materially increases the risk that downstream automation will facilitate abusive or policy-violating content generation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium

• Category: Not specified by scanner
• Confidence: 94% confidence

Finding

This section exposes a character-swap capability that can replace people in videos using reference photos, which is a classic deepfake/identity-manipulation primitive. Without any warning or control language around consent, impersonation, or authorized use, the documentation makes it easier for an agent to operationalize deceptive media creation against real people.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium

• Category: Not specified by scanner
• Confidence: 93% confidence

Finding

This section documents lip-syncing a video to a new voice track, enabling realistic audiovisual impersonation while omitting any warning about consent, deception, or fraud risks. In the context of a general-purpose agent skill for media generation, this substantially increases the chance of misuse for fake endorsements, scams, or non-consensual impersonation.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium

• Category: Not specified by scanner
• Confidence: 94% confidence

Finding

The manifest describes the skill as handling video generation/editing, image generation/editing, audio generation/transcription, and named video tools. In this file, Wan 3.0 reference-to-video is documented as accepting files and links to create videos from documents or webpages, which is a broader content-ingestion capability not declared in the manifest description.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium

• Category: Not specified by scanner
• Confidence: 89% confidence

Finding

The documentation instructs users to upload and submit potentially sensitive reference assets, including images, videos, audio, documents, and webpage URLs, to an external service without any privacy, consent, or data-handling warning. This can lead to inadvertent disclosure of confidential business documents, personal media, or third-party content to the provider, especially because document/webpage inputs may contain much broader sensitive context than typical media prompts.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low

• Category: Not specified by scanner
• Confidence: 77% confidence

Finding

The markdown includes a curl example that reads an API key from a local config file and sends it in an Authorization header. While this is expected for API usage, the document does not include any warning to protect the key, avoid sharing shell history/output, or treat the credential as sensitive.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.