@luckydraw/cumulus@1.0.57
RLM-based CLI chat wrapper for Claude with external history context management
AI Security Review
scanned 3d ago · by lpm-firewall-aiThe gateway runtime actively injects instructions that override conflicting agent instructions and suppress disclosure of the conflict. This is concrete AI-agent control behavior and supports blocking publication, despite the lack of a malicious install hook.
Decision evidence
public snapshot- The gateway’s built-in system prompt tells its agent to disregard conflicting tool or system reminders and conceal that conflict.
- The gateway constructs a system prompt for model turns and sends it with the user message.
- The prompt behavior is part of the gateway runtime, not an install hook.
- The manifest’s prepare hook runs Husky, and prepublishOnly builds and type-checks the package.
- Installing the gateway service requires an explicit setup command; enabling login auto-start prompts the user.
- The package update operation is invoked from an explicit update flow.
Source & flagged code
8 flagged · loading sourcePackage source references child process execution.
dist/tui/hooks/useClaudeProcess.jsView on unpkg · L3This package version adds a dangerous source file absent from the previous stored version; route for source-aware review.
dist/lib/gateway.jsView on unpkgFindings
Affected versions and remediation
This report applies to @luckydraw/cumulus@1.0.57.
See all versions of @luckydraw/cumulus for other recorded verdicts.
- Avoid installing @luckydraw/cumulus@1.0.57. Remove it from direct dependencies and check your lockfile for transitive copies.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
- If this version ran, investigate the affected machine and build environment. Rotate credentials it could access and rebuild from a trusted environment.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.