v1.0.0

Auto Shorts

BenignClawScan verdict for this skill. Analyzed Apr 30, 2026, 11:08 PM.

Analysis

Auto Shorts is a purpose-aligned cloud video-processing skill, but it uploads user media and prompts to NemoVideo and uses a service token/session, so users should trust that provider before uploading private content.

GuidanceBefore installing or using this skill, make sure you are comfortable sending your videos and editing instructions to the NemoVideo cloud service. Keep NEMO_TOKEN private, avoid uploading confidential media unless you trust the provider, and review generated clips before sharing them publicly.

Findings (7)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

Abnormal behavior control

Checks for instructions or behavior that redirect the agent, misuse tools, execute unexpected code, cascade across systems, exploit user trust, or continue outside the intended task.

Agent Goal Hijack

SeverityLowConfidenceHighStatusNote

SKILL.md

Backend says | You do ... "click [button]" / "点击" | Execute via API

The skill instructs the agent to translate backend GUI-style responses into API actions. This is purpose-aligned for the cloud editing workflow, but it means backend text can steer agent actions within the session.

User impactThe service backend may guide edits or exports through the agent rather than every step being explicitly initiated by the user.

RecommendationUse the skill for the specific video task you requested, and review summaries/downloads before relying on or publishing the output.

Tool Misuse and Exploitation

SeverityLowConfidenceHighStatusNote

SKILL.md

**Upload**: POST `/api/upload-video/nemo_agent/me/<sid>` — file: multipart `-F "files=@/path"`, or URL

The skill can upload user-provided media files or URLs and trigger cloud processing. This is central to the stated purpose, but it is still a meaningful external operation.

User impactVideos or media URLs provided to the skill will be sent to the NemoVideo cloud service for processing.

RecommendationOnly provide files or URLs you intend to process with this third-party cloud service.

Agentic Supply Chain Vulnerabilities

SeverityInfoConfidenceHighStatusNote

metadata

Source: unknown; Homepage: none

The skill is instruction-only and has no install-time code, but the registry metadata does not provide source or homepage provenance for the cloud integration.

User impactUsers have limited registry-provided information for independently verifying the provider behind the cloud API.

RecommendationReview the provider/domain and avoid uploading sensitive videos unless you trust the service.

Rogue Agents

SeverityLowConfidenceHighStatusNote

SKILL.md

closing the tab before completion orphans the job

The cloud render job can continue or become orphaned if the user leaves before completion. This is tied to the requested render workflow and not evidence of self-propagation, but users should know jobs may persist server-side.

User impactA started cloud render may keep running even if the local interaction is interrupted.

RecommendationStart exports only when you intend to let the cloud job finish, and check service state if a task is interrupted.

Permission boundary

Checks whether tool use, credentials, dependencies, identity, account access, or inter-agent boundaries are broader than the stated purpose.

Identity and Privilege Abuse

SeverityLowConfidenceHighStatusNote

SKILL.md

Every API call needs `Authorization: Bearer <NEMO_TOKEN>`

The skill requires a NemoVideo bearer token, and can obtain an anonymous token if one is not present. This credential use is expected for the service integration and no token leakage or unrelated use is shown.

User impactThe skill uses a NemoVideo token to create sessions, upload media, check credits, and render outputs.

RecommendationKeep NEMO_TOKEN private and revoke or rotate it if you no longer want the skill to access the NemoVideo service.

Sensitive data protection

Checks for exposed credentials, poisoned memory or context, unclear communication boundaries, or sensitive data that could leave the user's control.

Memory and Context Poisoning

SeverityLowConfidenceHighStatusNote

SKILL.md

Keep the returned `session_id` for all operations.

The skill maintains a cloud session and later polls session state for drafts, video info, and generated media. This is expected for video processing, but session state becomes trusted context for subsequent actions.

User impactInformation about the editing session and generated media may persist in the cloud session while the task is active.

RecommendationDo not include private or sensitive content unless you are comfortable with it being retained in the service session for processing.

Insecure Inter-Agent Communication

SeverityLowConfidenceHighStatusNote

SKILL.md

**Send message (SSE)**: POST `/run_sse` — body ... `"app_name":"nemo_agent"`

The skill sends user messages to a remote NemoVideo agent endpoint over SSE and processes tool-call style events internally. The endpoint is fixed and purpose-aligned, but user prompts and task context leave the local agent boundary.

User impactYour editing instructions and related session context are processed by a remote cloud agent/service.

RecommendationAvoid including confidential details in prompts unless they are necessary for the video edit and you trust the provider.