@livx.cc/agentx@0.99.103
Edge-native AI agent runtime — drives a virtual filesystem via any LLM (ai.libx.js). Same bytes run in node, browser, or edge.
AI Security Review
scanned 4d ago · by lpm-firewall-aiThe default Bodify request appears package-aligned, but an unrestricted BODIFY_URL override can redirect a saved vault credential, leaving a concrete conditional exposure risk.
Decision evidence
public snapshot- The package’s prepare script only sets a Git hooks path; its CLI entry point is dist/cli.js.
- The CLI can retrieve a saved Para vault token and use it as the fallback credential for Bodify secret retrieval.
- The request sends that credential as a bearer token to BODIFY_URL, which defaults to bodify.bod.ee and has no destination allowlist in the inspected code.
- CLI startup loads environment settings and calls the secret retrieval function when starting a run.
- The flagged CLI location builds a Claude worktree path; it does not establish agent configuration mutation.
- The flagged declaration describes agent definitions, while the flagged runtime code loads project instruction files into the prompt.
- The credential request defaults to bodify.bod.ee for the Bodify secret-fetch flow; the inspected source does not show a fixed unrelated recipient.
- The flagged agent-related locations show worktree path construction, an interface declaration, and project instruction loading, not automatic foreign agent configuration changes.
Source & flagged code
12 flagged · loading sourcePackage source references child process execution.
dist/tools.shell.jsView on unpkg · L891Findings
Affected versions and remediation
This report applies to @livx.cc/agentx@0.99.103.
See all versions of @livx.cc/agentx for other recorded verdicts.
- Review the evidence and your use of @livx.cc/agentx@0.99.103 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.