Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Sui Vibe
v1.0.1W.A.L.V.I.S. - AI-powered knowledge manager. Save anything from Telegram — links, text, images. Auto-tag and organize with AI; store on Walrus decentralized...
⭐ 0· 274·0 current·0 all-time
by@kuuga-0
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The name/description, the SKILL.md, and the included code (CLI, hook, fastpath plugin, web UI) are consistent with a Telegram knowledge manager that stores data locally under ~/.walvis and syncs to Walrus testnet. However the registry metadata says no required env vars while README/CLAUDE.md/bin/cli.js and SKILL.md expect an LLM API endpoint/key and optional Sui wallet info — a clear metadata mismatch. Also metadata.openclaw suggests installing an npm package 'walvis' while this repository contains the skill sources locally; that discrepancy is worth confirming before running 'npx walvis'.
Instruction Scope
Runtime instructions explicitly tell the agent to read and write files under ~/.walvis, to fetch pages using WebFetch and a real browser tool (open + snapshot + screenshot), and to upload screenshots via curl to walrus-testnet publisher endpoints. Those behaviors are coherent with a bookmarking tool, but they grant the agent broad network access (fetch arbitrary user-supplied URLs and upload binary screenshots to external endpoints) and instruct the agent to always prefer the browser tool — which increases the blast radius if misused.
Install Mechanism
There is no install spec in the registry entry, but the repository contains a CLI (bin/cli.js), package.json and package-lock.json and many JS/TS files — so installation will involve copying files and running npm in your environment or running 'npx walvis'. The code calls child_process (execSync/spawnSync) in places (installer detects Docker/OpenClaw) — expected for an installer but worth attention. No downloads from untrusted hosts are present in the manifest, but verify the npm package name and origin if you run 'npx walvis'.
Credentials
Registry metadata claims no required environment variables, yet the CLI and docs request an LLM API key/endpoint (llmEndpoint/llmApiKey / WALVIS_LLM_API_KEY) and optionally a Sui wallet address. The skill will prompt for and persist LLM credentials during setup and uses them for content analysis — requesting an LLM key is proportionate to the stated AI-analysis purpose, but the mismatch between declared requirements and actual code/installer is a red flag you should resolve. No unrelated secrets (AWS, GitHub, etc.) are requested.
Persistence & Privilege
The skill does not request 'always: true' and is user-invocable. It includes an OpenClaw message hook that rewrites messages (fast-path routing) and can auto-save bare URLs if manifest.autoSave is enabled; the fastpath plugin defaults to ON for new installs. This behavior (message rewriting/automatic saves and plugin that bypasses LLM for deterministic commands) is within the skill's purpose but is a privilege you should be conscious of because it changes incoming message content and can trigger automatic uploads/syncs.
What to consider before installing
What to check before installing:
- Confirm where you'll run the installer: the repo includes code and an installer (bin/cli.js); prefer reviewing/inspecting the files locally before running any npx install. Do not blindly run 'npx walvis' unless you trust the npm package owner.
- Expect to provide an LLM API key (llmApiKey / WALVIS_LLM_API_KEY). The registry claims no env vars, but the CLI and docs require this; ask the author how/where the key is stored (manifest or env) and avoid supplying high-privilege keys (use a scoped/test key if possible).
- Be aware the skill will read/write ~/.walvis (manifest, spaces, media) and may upload screenshots and space blobs to walrus-testnet endpoints. If you plan to keep data private, do not run `/walvis sync` or enable uploads until you audit the code and configuration.
- The skill instructs the agent to use a real browser tool to open arbitrary URLs and take snapshots — this is necessary for rich bookmarks but increases risk (drive-by content, JS execution, etc.). Run the skill in an isolated environment or container if you have sensitive data or are unsure.
- The OpenClaw hook rewrites incoming messages and can auto-save bare URLs — if you do not want automatic behavior, set manifest.autoSave = false and disable fastPath or review the hook before enabling.
- Review where the CLI/installer inspects Docker mounts and writes into OpenClaw config (it calls docker inspect / exec). If you run the installer on a machine with Docker, it may detect containers and host paths; verify these operations are acceptable in your environment.
- If you want to proceed, audit where credentials are stored (manifest.json, openclaw.json) and prefer testnet/demo keys initially. If unsure, run the web UI locally in dev mode and test with dummy data before connecting your real LLM key or syncing to Walrus.bin/cli.js:241
Shell command execution detected (child_process).
extensions/walvis-fastpath/index.js:66
Shell command execution detected (child_process).
scripts/smoke-walvis-scripts.mjs:13
Shell command execution detected (child_process).
extensions/walvis-fastpath/index.js:565
Environment variable access combined with network send.
scripts/smoke-fastpath-plugin.mjs:155
Environment variable access combined with network send.
skill/scripts/analyze.ts:77
Environment variable access combined with network send.
bin/cli.js:8
File read combined with network send (possible exfiltration).
extensions/walvis-fastpath/index.js:3
File read combined with network send (possible exfiltration).
scripts/smoke-fastpath-plugin.mjs:4
File read combined with network send (possible exfiltration).
skill/scripts/walrus-sync.ts:15
File read combined with network send (possible exfiltration).
Patterns worth reviewing
These patterns may indicate risky behavior. Check the VirusTotal and OpenClaw results above for context-aware analysis before installing.Like a lobster shell, security has layers — review code before you run it.
latestvk97665bsfd4afzc06d85dk9n198309e8
License
MIT-0
Free to use, modify, and redistribute. No attribution required.