Skillv1.0.0

ClawScan security

Video Prompt · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

SuspiciousApr 25, 2026, 9:17 AM

Verdict: suspicious
Confidence: medium
Model: gpt-5-mini
Summary: The skill's declared purpose (server-side AI video generation) matches most of its runtime instructions, but there are inconsistencies and a few instructions that ask the agent to read local paths/metadata and to upload potentially large user files to an external service — verify before installing or providing credentials.
Guidance: This skill appears to do what it claims (upload your text and media to a cloud rendering service and return videos), but a few things to check before installing or providing credentials: 1) Confirm the service domain (mega-api-prod.nemovideo.ai) and vendor trust/privacy policy — uploaded media (and any sensitive content) will be transmitted to that endpoint. 2) Ask the publisher to resolve the metadata mismatch: SKILL.md frontmatter lists ~/.config/nemovideo/ while the registry shows no configPaths — clarify whether the skill will read local config directories. 3) If you prefer not to expose a persistent NEMO_TOKEN, use the anonymous-token flow only for non-sensitive test content; anonymous tokens seem short-lived and limited. 4) If you need stronger assurance, request a minimal reproducible example of requests the skill will send (headers, exact endpoints) and confirm that attribution headers only read this skill's metadata and do not enumerate other installed skills. Provide these clarifications to increase confidence and reduce risk.

Review Dimensions

Purpose & Capability: noteSkill name/description (generate videos from text) align with the endpoints and flows in SKILL.md and the single required env var NEMO_TOKEN — that is expected. However the SKILL.md frontmatter lists a required config path (~/.config/nemovideo/) while the registry metadata reported no required config paths; this mismatch should be clarified.
Instruction Scope: concernSKILL.md instructs creating sessions, uploading user files (up to 500MB), streaming SSE, and polling export endpoints — all consistent with a cloud-render service. It also instructs reading this file's YAML frontmatter and detecting install path (~/.clawhub/, ~/.cursor/skills/) to populate X-Skill-Platform header. Those steps require reading local filesystem paths and could reveal environment/install details beyond the single declared env var; the skill's registry metadata did not consistently declare those config path reads.
Install Mechanism: okInstruction-only skill with no install spec and no code files — lowest-risk install mechanism. Nothing is written to disk by an installer.
Credentials: noteOnly NEMO_TOKEN is declared as required, which is proportionate for a remote video API. The skill also documents a fallback anonymous-token acquisition flow (POST to an external auth endpoint using a generated client UUID) that does not require user secrets. Still, uploading large user files to a third-party service is inherently sensitive — users should confirm they trust the endpoint before supplying videos or a persistent token.
Persistence & Privilege: okalways:false and no install scripts or claims of modifying other skills or global configs. The skill manages ephemeral sessions and tokens for remote rendering but does not request permanent agent-level privileges in the provided instructions.