registry / @rapchic/roast / 0.1.2

@rapchic/roast@0.1.2

Evidence-based code roast CLI — install roast skills for AI coding agents

AI Security Review

scanned 9d ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. A global npm postinstall mutates Cursor's agent-control directories when ~/.cursor exists. It also writes project-local .cursor assets in the install working directory.

Static reason

High-risk behavior combination matched malicious policy.

Trigger

npm install -g @rapchic/roast, unless ROAST_SKIP_POSTINSTALL=1 or ~/.cursor is absent.

Impact

Changes a foreign AI-agent control surface and can affect Cursor behavior without an explicit roast command.

Mechanism

unconsented lifecycle deployment of Cursor skills, commands, and rules

Policy narrative

On global installation, the postinstall hook detects ~/.cursor and immediately invokes the Cursor installer with automatic confirmation. That installer copies package-controlled skill, command, and rule files into Cursor's global control directories and writes project-local .cursor assets in the current directory. This is lifecycle-triggered agent control-surface mutation rather than an explicit user install command.

Rationale

Source inspection confirms an unconsented postinstall mutation of a foreign AI-agent control surface. The global-install and existing-Cursor guards reduce scope but do not make the mutation user-invoked.

Evidence

package.jsonscripts/postinstall.jsinstaller/src/install.jsinstaller/src/clients/cursor.jsinstaller/src/project.js~/.cursor/skills/roast~/.cursor/commands/roast*.md~/.cursor/rules/roast.mdc.cursor/skills/roast

@rapchic/roast@0.1.2

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code