Disclosed weakness · no CVE recorded
PraisonAI corrected insecure server defaults and agent trust boundaries
PraisonAI · PraisonAI Python and TypeScript packages
First documented disclosure · Review checkpoint
Summary
A grouped maintainer advisory describes insecure server defaults and gaps in agent permissions, file and network access, project trust and input handling.
Severity and exploitation evidence
Critical · 9.8/10 · CVSS 3.1. Severity source.
Exploitation not established: The reviewed sources establish the weakness or a controlled demonstration; malicious exploitation against users is not established.
Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.
How the weakness works
Controls were absent or inconsistent across server routes, tools, project settings and sandbox backends; the release introduces shared validation and stronger defaults.
Conditions required
Exposure depends on the component and configuration: reachable unauthenticated servers, an untrusted project folder, attacker-influenced tool input, or a policy the selected backend cannot enforce.
Potential impact
Depending on the area, unauthorized agent-tool use, file access, internal network requests, credential exposure or unintended execution.
Affected versions and configurations
praisonai < 4.8.0; praisonaiagents < 1.8.0; Python praisonai-code/bot/mcp/sandbox/browser/deploy < 0.2.0; npm praisonai < 1.8.0.
Fix and mitigation status
The advisory names praisonai 4.8.0, praisonaiagents and npm praisonai 1.8.0, and the six listed Python component packages 0.2.0. These are issue-specific reported corrections, with release-note limits.
Practical next steps
Upgrade every installed affected package, review changed defaults and opt-outs, keep public servers authenticated, and use a backend that can enforce the requested sandbox policy.
Operating lesson
A security setting needs consistent enforcement across every route, tool and child process.
Evidence and limits
This is one grouped advisory, not a count of individual defects or incidents. The maintainer’s Critical/CVSS 3.1 9.8 score applies to the highest-scored default server exposure; other areas have lower or configuration-dependent ratings. The release notes retain a crawl4ai redirect limit, dashboard/replay token-setting work in progress and a file-search default in progress. npm 1.8.0 and the two main Python versions were published before this review cutoff; all component implementations were not independently audited. Prior separately numbered PraisonAI issues are not re-dated by this release. The praisonaiagents 1.8.0 source archive matched PyPI SHA-256; its AgentServer defaults to 127.0.0.1 and calls require_bind_token before starting. This is a focused source check, not an audit of every component.
Sources
- PraisonAI grouped security advisory
Primary source ·
- PraisonAI 4.8.0 changes and migration limits
Primary source ·
- npm PraisonAI publication metadata
Primary source ·
- PraisonAI 4.8.0 PyPI metadata
Primary source ·