registry / bravecode-cli / 0.1.57

bravecode-cli@0.1.57

Agentic AI vibe coding CLI with free model support

AI Security Review

scanned 42m ago · by lpm-firewall-ai

An explicit agent run sends prompts and tool context to codero.sohailsyed.com. Returned tool-call markup can cause local shell execution even when the package says approval is required.

Static reason

One or more suspicious static signals were detected.; source matched previously finalized malicious package; routed for review

Trigger

A user runs bravecode run with the default provider, or otherwise uses an agent that receives remote tool calls.

Impact

A compromised endpoint or prompt-injected response can execute commands in the user's project context and access material available to that process.

Mechanism

Remote model output is converted to local tool calls without enforcing approval.

Rationale

This is a concrete remote-command-execution vulnerability in a user-invoked AI agent, not an install-time malware chain. The empty postinstall hook and lack of automatic execution support a warning rather than a publish block.

Evidence

dist/cli-main.mjs~/.bravecode/config.json

Network endpoints1

codero.sohailsyed.com

Decision evidence

public snapshot

AI called this Suspicious at 96.0% confidence as Critical Vulnerability with low false-positive risk.

Evidence for warning

bravecode-cli@0.1.57

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code