Back to skill
Skillv1.0.0

ClawScan security

Subtitle Translator Free · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

SuspiciousApr 28, 2026, 5:00 PM
Verdict
suspicious
Confidence
medium
Model
gpt-5-mini
Summary
The skill mostly does what it says (upload video, request a token, call a subtitle-translation backend), but there are inconsistencies and a few behaviors that increase privacy/risk (reading local paths/skill frontmatter, auto-obtaining tokens, unknown source) that you should understand before installing.
Guidance
Before installing or using this skill: 1) Confirm the backend domain (mega-api-prod.nemovideo.ai) and owner are trustworthy—there is no homepage or provenance listed. 2) Understand that videos and derived subtitles are uploaded to that external service; check its privacy/retention policy if you have sensitive content. 3) The skill will read local paths (the skill frontmatter and may inspect ~/.config/nemovideo/ and install paths) — ensure those directories don't contain secrets you don't want shared. 4) Prefer to provide your own NEMO_TOKEN (if you have one) rather than allowing the skill to auto-create an anonymous token; ask what data that anonymous token ties to and how long uploads/outputs are retained. 5) Ask the publisher to resolve the metadata inconsistency (registry vs SKILL.md configPaths) and to provide a homepage/privacy policy. If you can't verify the service or you have sensitive video/audio, treat this skill as higher risk and avoid installing it until provenance and data-handling are clarified.

Review Dimensions

Purpose & Capability
noteThe declared purpose (translate subtitles and render video) matches the instructions to upload files and call a remote rendering API. However the SKILL.md frontmatter includes a configPaths entry (~/.config/nemovideo/) that wasn't present in the registry summary, which is an inconsistency worth questioning: why would a simple translator need to read that path?
Instruction Scope
concernRuntime instructions tell the agent to: (a) read this file's YAML frontmatter at runtime, (b) detect install path (~/.clawhub, ~/.cursor/skills) to set an attribution header, and (c) upload user video files to an external backend and poll SSE endpoints. Reading the install path and a local config dir means the agent will access the user's filesystem beyond just the uploaded video. The skill also instructs generating an anonymous token via a POST to an external endpoint if NEMO_TOKEN is not present — that network action is expected for a cloud service but should be explicit to the user because it creates/uses credentials and transmits files off-device.
Install Mechanism
okNo install spec or code files — instruction-only skill. This limits on-disk execution risk (no downloaded archives or third-party installs).
Credentials
noteThe only declared required env var is NEMO_TOKEN (primary credential), which is appropriate for a cloud API. But the instructions also describe auto-obtaining an anonymous token if none is present and reference reading ~/.config/nemovideo/ and install paths to set headers. Those filesystem accesses could surface other credentials or config, so the requested environment access is small but the implied filesystem access increases exposure.
Persistence & Privilege
okalways:false and no requests to change other skills or global agent settings. The skill will run network calls and may be invoked autonomously (normal for skills), but it doesn't request elevated platform privileges.