thalamus@1.5.86
⚠ Under reviewMobile and Web client for Claude Code
AI Security Review
scanned 19d ago · by lpm-firewall-aiThe inspected runtime paths did not establish credential theft or agent-control abuse. The install hook extracts an archive whose expanded contents remain materially unresolved, so manual review is appropriate.
Decision evidence
public snapshot- The postinstall hook runs the local tool unpacker.
- The unpacker selects a platform-specific archive and extracts it into tools/unpacked.
- The offline analysis identifies the flagged archive as gzip but says its expanded contents exceed the inspection limit.
- The Anthropic OAuth token registration path sends the token to the configured Thalamus server after the user runs the connect flow.
- The reviewed OAuth path uses the configured Thalamus server and is activated by an explicit connection flow.
- The flagged agent-configuration locations inspected contain transcript and friction-metrics logic, not an agent configuration mutation.
Source & flagged code
20 flaggedPackage defines install-time lifecycle scripts.
package.jsonView on unpkgPackage source references child process execution.
bin/thalamus-mcp.mjsView on unpkg · L2Findings
Affected versions and remediation
This report applies to thalamus@1.5.86.
See all versions of thalamus for other recorded verdicts.
- Review the evidence and your use of thalamus@1.5.86 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.