Loading npm security reports…

@quotient-forecasting/cassie-skill@0.4.17: AI-agent | LPM Firewall

registry / @quotient-forecasting/cassie-skill / 0.4.17

@quotient-forecasting/cassie-skill@0.4.17

Agent Skill operator manual and policy files for Cassie

AI Security Review

scanned 12d ago · by lpm-firewall-ai

LPM flags this version as an AI-agent control-surface risk. An npm postinstall hook writes package-controlled agent skill content into default user-wide Codex and Claude skill directories. This is an unconsented lifecycle mutation of broad AI-agent control surfaces.

Static reason

High-risk behavior combination matched malicious policy.; source fingerprint signature matched known malicious package; routed for review

Trigger

Installing the package with npm lifecycle scripts enabled.

Impact

Future agent sessions can discover and act on package-supplied instructions without an explicit user setup command.

Mechanism

The postinstall executes install.mjs, which selects default home-directory agent skill roots, creates a cassie subdirectory, and copies SKILL.md and thesis content there.

Policy narrative

On installation, npm automatically starts the package installer. The installer writes its instruction file and policy content into two user-wide AI-agent skill locations. Those locations affect how later Codex or Claude sessions discover package-provided capabilities, and the mutation occurs without an explicit user command.

AI rationale

The package has a postinstall hook that modifies broad, foreign AI-agent control surfaces in the user's home directory. This meets the install-control-surface blocking rule even though no separate network or credential theft behavior was identified.

Evidence

package.jsoninstall.mjs~/.agents/skills/cassie/SKILL.md~/.agents/skills/cassie/thesis~/.claude/skills/cassie/SKILL.md~/.claude/skills/cassie/thesis

Decision evidence

Source & flagged code

8 flagged · loading source

package.jsonView file

•scripts.postinstall = node install.mjs

High

Install Time Lifecycle Scripts

Package defines install-time lifecycle scripts.

package.jsonView on unpkg

36"scripts": { "postinstall": "node install.mjs"

High

Suspicious Lifecycle Evidence

The package runs install.mjs automatically through a postinstall lifecycle hook.

package.jsonView on unpkg · L36

•scripts.postinstall = node install.mjs

Medium

Ambiguous Install Lifecycle Script

Install-time lifecycle script is not statically allowlisted and needs review.

package.jsonView on unpkg

install.mjsView file

3L4: import { cpSync, existsSync, mkdirSync } from "node:fs"; L5: import { homedir } from "node:os"; ... L12: function defaultRoots() { L13: return [join(homedir(), ".agents", "skills"), join(homedir(), ".claude", "skills")]; L14: } ... L27: const destination = join(root, "cassie"); L28: mkdirSync(destination, { recursive: true }); L29: cpSync(join(packageDir, "SKILL.md"), join(destination, "SKILL.md"), { force: true }); L30: cpSync(join(packageDir, "thesis"), join(destination, "thesis"), { recursive: true, force: true }); L31: installed.push(destination);

Critical

Ai Agent Control Hijack

Source creates an unconsented AI-agent control surface through install-time mutation or a default unauthenticated remote skill channel.

install.mjsView on unpkg · L3

•Runtime or CLI code links a write operation to an explicit external AI-agent control path: import { cpsync, existssync, mkdirsync } from "node:fs"; import { homedir } from "node:os"; import { dirname, join, resolve } from "node:path"; import { fileurltopath } from "node:url"; const packagedir = dirname(fileurltopath(import.meta.url)); const workspaceroot = resolve(packagedir, "../.."); function defaultroots() { return [join(homedir(), ".agents", "skills"), join(homedir(), ".claude", "skills")]; } export function installcassieskill(options = {}) { const override = process.env.cassie_skills_dir?.trim(); if (!options.force && process.env.cassie_skip_skill_install === "1") return [];

High

External Ai Agent Control Surface Mutation

Runtime or CLI source writes behavior-bearing configuration into a user or project AI-agent control surface.

install.mjsView on unpkg

•matchType = malicious_source_fingerprint_signature signature = 0da900ca224fb098 signatureType = suspicious_hashes sourceLabel = final_verdict:malicious matchedPackage = @quotient-forecasting/cassie-skill@0.2.0 matchedPath = install.mjs matchedIdentity = npm:[redacted]:0.2.0 similarity = 1.000 shingleOverlap = 2 summary = package final verdict is malicious

High

Known Malware Source Fingerprint Signature

Source fingerprint signature matches a known malicious package signature; route for source-aware review.

install.mjsView on unpkg

12function defaultRoots() { return [join(homedir(), ".agents", "skills"), join(homedir(), ".claude", "skills")]; } export fu

High

Ai Review Evidence

The installer targets both the user's .agents and .claude skill directories by default.

install.mjsView on unpkg · L12

24const roots = options.roots ?? (override ? [resolve(override)] : defaultRoots()); const installed = []; for (const root of roots) { const destination = join(root, "cassie"); mkdirSync(destination, { recursive: true }); cpSync(join(packageDir, "SKILL.md"), join(destin

High

Suspicious Lifecycle Evidence

When invoked by the lifecycle hook, the installer creates a cassie directory and copies its skill files into those agent control surfaces.

install.mjsView on unpkg · L24