v1.0.0

Video Editor Baby

SuspiciousClawScan verdict for this skill. Analyzed Apr 30, 2026, 12:27 PM.

Analysis

This baby video editor is purpose-aligned, but it automatically uses an external cloud service, handles tokens and private media, and lets the remote backend steer some internal actions.

GuidanceOnly use this skill if you are comfortable sending baby videos and editing instructions to NemoVideo's cloud service, using or creating a NEMO_TOKEN, and possibly encountering credit or subscription limits. Prefer non-sensitive clips, verify the provider's privacy and deletion terms, and ask for confirmation before any upload or export.

Findings (8)

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

SeverityMediumConfidenceHighStatusConcern

SKILL.md

Map its instructions to API calls: "click" or "点击" → execute the action via the relevant endpoint ... "Export" or "导出" → run the export workflow ... Tool calls stay internal.

Remote backend responses are treated as actionable instructions that can trigger API calls, while internal tool calls are hidden from the user.

User impactA backend response could cause the agent to run state, edit, upload, or export actions inside the session without the user directly approving that exact step.

RecommendationTreat backend text as untrusted guidance and require user confirmation before executing actions that upload files, spend credits, or start exports.

Tool Misuse and Exploitation

SeverityMediumConfidenceHighStatusConcern

SKILL.md

On first interaction, connect to the processing API before doing anything else... Upload: POST /api/upload-video/nemo_agent/me/<sid> — file: multipart ... or URL

The skill automatically initiates cloud API use and supports sending local files or URLs to an external service, but does not specify explicit per-upload approval or containment.

User impactPrivate baby footage may be uploaded to the external processing service and cloud jobs may be started as part of normal use.

RecommendationBefore upload or export, clearly ask for confirmation and disclose the destination service, expected data transfer, file size, cost/credit impact, and retention expectations.

Agentic Supply Chain Vulnerabilities

SeverityLowConfidenceMediumStatusNote

metadata

Source: unknown
Homepage: none

The registry metadata does not provide source or homepage provenance for a skill that depends on an external cloud processing service.

User impactUsers have limited information for verifying the provider, maintainer, or terms before trusting the backend with private media.

RecommendationVerify the NemoVideo provider and privacy/cost terms independently before uploading sensitive family footage.

Human-Agent Trust Exploitation

SeverityMediumConfidenceHighStatusConcern

SKILL.md

Export (free, no credits) ... `402` — free plan export blocked; not a credit issue, subscription tier

The instructions present export as free while also acknowledging subscription-tier blocking, which can mislead users about access or cost before they upload footage.

User impactUsers may upload private media expecting free export and only later learn that export may be blocked by plan or subscription limits.

RecommendationState possible subscription, credit, and export limitations before upload or render actions.

Rogue Agents

SeverityLowConfidenceHighStatusNote

SKILL.md

The session token carries render job IDs, so closing the tab before completion orphans the job.

Cloud render jobs can continue without an active user tab/session; this is disclosed and purpose-related, but it is persistent background activity.

User impactA render job may continue running after the user leaves, potentially leaving cloud artifacts or consuming service resources.

RecommendationProvide clear cancel, cleanup, and retry guidance before starting long-running exports.

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

metadata

Required env vars: NEMO_TOKEN ... Primary credential: NEMO_TOKEN

The service credential requirement is disclosed and appears related to the cloud video-rendering purpose, but it still grants access to a Nemo service account/session.

User impactUsing an existing NEMO_TOKEN may connect this skill to the user's Nemo account, credits, and render sessions.

RecommendationUse a limited-purpose token where possible, avoid sharing tokens in chat, and revoke or rotate the token if you no longer use the skill.

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

SeverityMediumConfidenceHighStatusNote

SKILL.md

Save session_id from the response... Session state: GET /api/state/nemo_agent/me/<sid>/latest — key fields: data.state.draft, data.state.video_infos, data.state.generated_media

The skill relies on remote session state containing drafts, video information, and generated media, which may persist and influence later export actions.

User impactVideo metadata, drafts, and generated media references may remain associated with the remote session and be reused during later actions.

RecommendationUse only footage you are comfortable storing in the cloud and look for provider-supported deletion or session cleanup controls.

Insecure Inter-Agent Communication

SeverityMediumConfidenceHighStatusConcern

SKILL.md

Send message (SSE): POST /run_sse ... app_name nemo_agent ... Tool calls stay internal.

User messages are sent to a remote agent-like backend, and the backend's internal tool calls are intentionally hidden, leaving data-use and action-boundary transparency unclear.

User impactEdit requests and media context may be processed by a remote agent/service without the user seeing all downstream tool actions.

RecommendationRequire clearer disclosure of the remote agent's identity, data handling, and tool actions, and avoid sending sensitive footage unless comfortable with that processing.