registry / glyph-compress / 1.21.2

glyph-compress@1.21.2

Semantic compression for IDE↔LLM communication using radical glyphs — CLI, MCP server, and VS Code extension, with a Context Router, team-shared dictionaries, and compression calibrated against real provider tokenizers.

AI Security Review

scanned 1d ago · by lpm-firewall-ai

LPM treats this as warn-only first-party agent extension lifecycle risk. A VS Code extension can modify AI instruction files in the active workspace when its opt-in setting is enabled. Its explicit local proxy forwards compressed LLM requests to the user-configured upstream.

Static reason

No blocking static signals were detected.; previous stored version diff introduced dangerous source

Trigger

User enables glyphCompress.autoUpdateWorkspaceRules and activates VS Code, or explicitly starts the proxy.

Impact

Changes Cursor/Copilot instruction files and can relay LLM request content and headers to the configured endpoint.

Mechanism

Guarded AI-workspace instruction mutation and user-invoked HTTP proxying.

Rationale

This is not malicious because no lifecycle hook or stealth/exfiltration chain was found. Warn for the guarded extension lifecycle behavior that writes AI-agent instruction files.

Evidence

package.jsonvscode-ext/package.jsonvscode-ext/extension.jssrc/proxy.jsbin/mcp-server.js.cursorrules.github/copilot-instructions.md

Network endpoints1

api.openai.com

Decision evidence

public snapshot

glyph-compress@1.21.2

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code