Omnibot Skills

• Outcome: Review
• Latest audit: Jul 22, 2026 · 4:55 AM UTC
• Version: 0.1.3

Overview

This browser-automation skill is mostly coherent, but it gives agents broad live-browser authority and includes under-scoped guidance for captcha interaction, real logged-in sessions, clipboard access, network evidence, and publishing actions.

Install only if you intentionally want an agent to operate a real logged-in Chromium browser. Use a separate browser profile or test account where possible, avoid sensitive banking/payment/admin sessions, review network or clipboard outputs before sharing them, and require explicit human confirmation before posting, submitting, purchasing, or attempting any captcha/human-verification workflow.

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (17)

Context-Inappropriate Capability

Medium

• Confidence: 94% confidence
• Finding: The command reference documents an explicit workflow for inspecting and operationalizing captcha challenges, including coordinate mapping and agent-guided drag/click steps. Even though it states the feature is experimental and not production-ready, this still expands the skill into anti-bot challenge interaction that is outside a normal browser inspection/automation need and can facilitate automated or semi-automated bypass of access controls.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The README explicitly promotes operation in a real browser with real cookies, real extensions, and real user sessions, but does not prominently warn that an agent acting in such a context can access authenticated data and perform actions as the user. In a skill meant for AI agents, this omission is security-relevant because it normalizes high-trust automation without clear consent, scope limits, or privacy/account-impact cautions.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: The examples showcase automated login, authenticated data extraction, court-record analysis, and one-click publishing, but they omit clear warnings about irreversible actions, privacy exposure, and compliance risks. In the context of an agent skill for live browser control, these examples could encourage operators to let agents act on production accounts or sensitive data without sufficient review or guardrails.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The README explicitly emphasizes use of a real browser with real cookies, extensions, and user sessions, which means any agent action occurs in an authenticated context with access to sensitive data and account capabilities. Without prominent warnings and consent boundaries, users may underestimate the privacy, security, and account-impact risks of letting an agent operate inside their live session.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The showcased workflows include auto-login, extraction of backend/account data, court-record analysis, and automatic publishing to third-party platforms, all of which can involve sensitive information and irreversible actions. Presenting these examples without explicit warnings, approval requirements, or safe-use guidance increases the risk of unauthorized data access, accidental posting, privacy violations, or harmful agent behavior in real user accounts.

Missing User Warnings

Medium

• Confidence: 86% confidence
• Finding: El texto promociona acceso a un navegador real con cookies, extensiones y sesiones de usuario reales, lo que implica capacidad para leer datos sensibles y ejecutar acciones con efectos reales. Al no advertir explícitamente sobre privacidad, consentimiento y acciones irreversibles, aumenta el riesgo de que un usuario o agente use la habilidad en cuentas reales sin entender el alcance ni las consecuencias.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: Las demostraciones describen inicio de sesión automático, extracción de datos analíticos, lectura de documentos judiciales y publicación automática en plataformas reales, pero no incluyen advertencias sobre autorización, privacidad, términos de servicio ni riesgo de daño irreversible. En el contexto de una skill para agentes autónomos con control del navegador, estos ejemplos normalizan operaciones sensibles sobre cuentas y contenido reales, lo que eleva significativamente el riesgo de uso inseguro o abusivo.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The README explicitly promotes use of a real Chromium session with real cookies, extensions, and user sessions, which materially increases the risk of exposing authenticated data, account state, console logs, and network traffic to an agent. In a browser-automation skill, failing to warn about privacy, credential, and account-safety implications can lead users to run the tool against sensitive logged-in sessions without informed consent or safe operating guidance.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The documentation advertises screenshots, full-page captures, console logs, network logs, and API evidence collection, all of which can contain tokens, personal data, internal URLs, request payloads, or other secrets. In the context of an agent skill designed to inspect live browser state, omitting a warning about sensitive-data capture makes accidental over-collection and disclosure substantially more likely.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The automated publishing example shows the agent filling content, configuring monetization-related settings, previewing, and publishing on behalf of the user, but it does not clearly warn that the tool can submit irreversible actions using the user's authenticated account. In this skill context, that omission raises the risk of unauthorized posting, accidental submission, reputational harm, or unintended business/account changes.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The README prominently demonstrates high-impact browser automation use cases such as automatic login, backend data extraction, and one-click publishing, but does not provide corresponding user-facing warnings, consent requirements, or boundaries for handling authenticated sessions and sensitive data. Because this skill operates against a real Chromium profile with real cookies and user sessions, the omission increases the risk that agents or users will perform privacy-invasive or account-impacting actions without appreciating the consequences.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The README promotes automation against a real logged-in Chromium session, including reading pages, clicking buttons, filling forms, extracting content, and inspecting network traffic, but it does not clearly warn users that this can expose cookies, private data, account contents, and cause irreversible account actions such as posting or form submission. In this skill context, the omission is more dangerous because the tool is explicitly designed to operate on real user sessions rather than isolated test browsers, increasing privacy, compliance, and account-integrity risk.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The README explicitly promotes access to real browser sessions, cookies, extensions, network logs, and live user state, but does not present corresponding warnings about credential exposure, session hijacking risk, sensitive data capture, or unintended actions on authenticated accounts. In the context of an agent skill that can drive a real browser, this omission materially increases the chance that operators will use it against personal or production sessions without understanding the privacy and security consequences.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The showcased workflows include automated login, backend data extraction, legal-document review, social-media browsing, and one-click content publishing, all of which can trigger sensitive reads or irreversible writes on real user accounts. Because the examples normalize high-impact account actions without nearby cautions, approval gates, or warnings about verifying targets and consequences, they make unsafe deployment more likely in exactly the scenarios where mistakes are costly.

Missing User Warnings

Medium

• Confidence: 88% confidence
• Finding: The skill explicitly permits `clipboard write` and describes operational use without requiring user awareness or consent, even though clipboard contents are a shared user resource and overwriting them can disrupt user workflows or replace sensitive copied data. In a browser automation skill, this is more dangerous because agents may execute commands directly and repeatedly, making silent clipboard modification a realistic side effect rather than a theoretical capability.

Missing User Warnings

Low

• Confidence: 80% confidence
• Finding: The skill exposes `browser notify <title> [message]` as a supported action but does not warn that this will visibly display a notification to the user. While lower impact than clipboard modification, unexpected notifications can confuse users, leak task context on-screen, or create social-engineering opportunities if an agent sends deceptive prompts.

Missing User Warnings

Medium

• Confidence: 88% confidence
• Finding: Clipboard read/write capability can expose passwords, tokens, personal data, and other sensitive user content from the browser environment, yet the reference presents it as a routine capability without privacy or consent warnings. In an agent skill, normalizing clipboard access increases the risk of silent data collection or overwriting user clipboard contents during automation.

VirusTotal

VirusTotal findings are pending for this skill version.

Static analysis

No suspicious patterns detected.