Security audit

Brainstorming

Security checks across malware telemetry and agentic risk

Overview

This skill is coherent and openly documented, but it automatically processes every completed HiJavis voice or keyboard unit and can create pending cards and chat summaries without a pre-run user confirmation.

Install only if you are comfortable with an enabled HiJavis skill examining every completed voice or keyboard unit to decide whether to create a brainstorm card. The behavior is disclosed and purpose-aligned, but it is broad: sensitive conversations could be processed and summarized into pending cards or chat digests before you discard them.

SkillSpector

By NVIDIA

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (8)

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The invocation examples include very broad natural-language triggers such as "help me organize my thoughts on this," which can easily overlap with ordinary conversation and cause the skill to activate unintentionally. In a skill that scans recent recordings and creates downstream artifacts, accidental triggering can lead to unwanted processing of sensitive speech and unintended creation of cards or chat summaries.

Missing User Warnings

High

Confidence: 97% confidence
Finding: The README states that the app listens to conversations, reads back through recent recordings, and posts summaries into chat, but it does not provide a prominent privacy warning or explain retention, consent, or sensitive-data handling. This is dangerous because users may not realize that ambient or recorded speech is being analyzed and transformed into stored artifacts, increasing the risk of privacy violations and inadvertent capture of confidential information.

Vague Triggers

Medium

Confidence: 86% confidence
Finding: The trigger set includes broad natural-language phrases like "brainstorm this" and generic ideation wording, which can cause the skill to activate in situations where the user did not clearly intend persistent card creation. Because activation leads to transcript retrieval and server-side writes, accidental invocation can expose or process more user content than expected.

Vague Triggers

High

Confidence: 97% confidence
Finding: The skill is explicitly designed to run on every completed voice/keyboard unit via dispatcher, with the agent deciding post hoc whether to act. That creates an overly broad activation boundary: every user transcript is processed by this skill, increasing privacy exposure and the chance of unwanted card creation or downstream handoff based on incidental conversation content.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The manual-use examples include open-ended phrases such as "help me organize my thoughts" and "turn this into a deck," which are common conversational expressions and may be matched too aggressively. In this skill, a false match is more than cosmetic because it can trigger transcript fetching, composition, and creation of a pending card in persistent storage.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The manifest states that the dispatcher invokes this skill directly for every completed unit, with no classifier or route matching, and lets the skill decide relevance itself. This creates an overly broad automatic invocation surface that can cause unintended processing of unrelated user content, increase exposure of transcripts to the skill pipeline, and generate unsolicited tasks or notifications from ordinary activity.

Vague Triggers

Medium

Confidence: 84% confidence
Finding: The trigger list includes generic phrases such as 'brainstorm this', which may plausibly appear in normal conversation and accidentally activate the skill. In this skill's context, accidental triggering is somewhat mitigated by the downstream Confirm/Discard human gate, but it can still cause unnecessary transcript access, card creation, and user-facing notification noise.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The `matches` string includes broad natural-language phrases such as `brainstorm`, `help me organize my thoughts`, and `turn this into a brief`, which can overlap with ordinary user speech and cause over-routing of unrelated content into the brainstorm skill. Because routing is automatic and driven by classifier/deliverable matching, overly general triggers increase the chance of unintended invocation, surprise data processing, and noisy or incorrect to-do creation.

VirusTotal

58/58 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.dangerous_exec, suspicious.env_credential_access

Shell command execution detected (child_process).

Critical

Code: suspicious.dangerous_exec
Location: test/data.test.js:17

Environment variable access combined with network send.

Critical

Code: suspicious.env_credential_access
Location: scripts/brainstorming.js:74