Back to skill

Security audit

GSD Headless (→ gsd-orchestrator)

Security checks for vulnerabilities and agentic risk

Overview

This appears to be a redirect/name-reservation skill, but installing it can also pull in an unpinned Node package, so users should review it before use.

Install gsd-orchestrator directly as the skill instructs. Treat gsd-headless as requiring review unless the publisher explains and pins the gsd-pi dependency or removes the install metadata from this redirect-only package.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:9
Finding

Unpinned Third-Party Package Installation in a Redirect-Only Skill

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 9–13
Vulnerability Type: Unpinned third-party Node.js dependency
Risk Level: Medium

Vulnerable Code:

yaml
requires:
  bins: [gsd]
install:
  kind: node
  package: gsd-pi
  bins: [gsd]

Technical Analysis

The skill metadata permits installation of the third-party Node.js package gsd-pi without specifying an exact version, integrity digest, or otherwise verifiable source. Dependency resolution can therefore select a package release that differs from the version reviewed by the skill publisher or auditor.

This creates a supply-chain risk because Node.js package installation can invoke package lifecycle scripts and install executable content. If the package registry account, package release process, or resolved package version is compromised, attacker-controlled code could run during installation with the privileges of the user or service installing the skill.

The dependency is also inconsistent with the skill's stated purpose. The document describes itself as a name reservation and directs users to install gsd-orchestrator; consequently, automatically provisioning gsd-pi expands the trusted computing base without an evident functional requirement.

Attack Path

  1. A user or automation system installs or processes the gsd-headless skill.
  2. The skill manager observes that the required gsd binary is unavailable.
  3. The unpinned gsd-pi package is resolved from the configured Node.js package source.
  4. An unexpectedly changed or compromised package release is downloaded.
  5. Package lifecycle scripts or installed executable code run with the privileges of the installing process.
  6. Malicious package behavior could then access resources available to that account.

Exploitation depends on compromise or malicious modification of the dependency or its distribution channel; no malicious payload was present in the audited pr ...[truncated 608 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove the requires and install metadata from this skill if it is solely a name reservation and redirect.
  2. If gsd-pi is genuinely required, pin it to a reviewed exact version rather than allowing unconstrained resolution.
  3. Enforce integrity verification through a trusted lockfile, package digest, or equivalent mechanism supported by the installation framework.
  4. Disable or restrict package lifecycle scripts where operationally possible.
  5. Document the package's purpose, trusted registry, publisher, reviewed version, and update procedure.
  6. Perform dependency installation with a least-privileged account and isolate it from sensitive credentials and host resources.
  7. Continuously monitor the dependency for ownership changes, compromised releases, known vulnerabilities, and unexpected lifecycle-script modifications.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep