registry / trimprompt / 1.0.36

trimprompt@1.0.36

TrimPrompt — Save 80% on LLM tokens. Zero-config CLI proxy for AI coding agents

AI Security Review

scanned 20h ago · by lpm-firewall-ai

LPM blocks this version under the AI-agent control-surface policy. Installation silently persists command-interception shims through user shell profiles and PATH. The shims alter common command execution in AI-agent environments; runtime telemetry can identify the developer.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

npm postinstall; later shell commands with AI-agent environment variables set

Impact

Unconsented interception of developer/agent command execution and telemetry exfiltration.

Mechanism

postinstall PATH/profile persistence and AI-agent command shimming

Policy narrative

The postinstall hook automatically invokes shim installation. It creates a user-level shim directory, prepends it through shell profile/PATH changes, and installs wrappers for common development commands. In detected AI-agent environments, those wrappers route commands through the package, creating an unconsented broad control point. Separately, runtime tracking can POST developer identity and aggregate activity to a hard-coded Railway endpoint.

Rationale

Source inspection confirms an unconsented postinstall mutation of broad shell command control surfaces intended to activate around AI agents. This meets the install-control-surface block boundary.

Evidence

package.jsoncli.jsshims.jssync.jsfile-watcher.jshooks/claude-hook.js~/.trimprompt/shims/*~/.bashrc~/.zshrc~/.profile~/.claude/hooks/post_tool_use.js

trimprompt@1.0.36

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code