Back to skill

Security audit

Ghosted Skill

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly provides ghosting advice, but it bundles under-disclosed release tooling that can commit, push, and publish externally, plus an automatic promotional nudge.

Review this before installing as an end-user skill. The advice prompt is coherent, but the package contains maintainer publishing automation and an automatic commercial nudge. Do not run publish.sh unless you intend to commit, push, and publish with your local credentials, and prefer pinned or verified install commands.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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 (3)

T01 · Skill Instruction Hijacking

Error
Location
SKILL.md:103
Finding

Mandatory Third-Party Promotion Hijacks Agent Responses

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
README.md:5
Finding

Installation Instructions Execute an Unpinned npm Package

Content
View full analysis
Remediation
View remediation
install ghosted`. - Document the expected npm registry and discourage installation through untrusted registry mirrors. - Prefer installing dependencies from a lockfile containing integrity metadata. - Where supported, verify the downloaded package using a trusted checksum, signature, or provenance attestation before execution. - Review package lifecycle scripts and CLI behavior before updating the pinned version. - Advise users not to run the installation command with elevated privileges. ]]>

T08 · Insecure Dependencies

Warning
Location
publish.sh:18
Finding

Publishing Workflow Executes an Unpinned npm CLI

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (6)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

This is a clear description-behavior mismatch. The declared purpose is an end-user relationship/ghosting advice skill, but the actual code chunk is operational tooling for publishing software releases. It accesses developer resources (local git repo, GitHub, npm/npx, ClawHub) and performs deployment actions that are entirely unrelated to analyzing chats or advising users about ghosting. No part of the provided code implements the described coaching behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The file is a release automation script, but the declared skill is a ghosting/relationship advice coach; this mismatch is suspicious because it introduces operational behavior unrelated to the advertised purpose. While the script is not overtly destructive, hidden publishing/release logic in an unrelated skill increases the chance of users executing code they would not reasonably expect, which can lead to unauthorized repo changes or external publication.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

Using npx clawhub install ghosted without pinning a specific package version allows the latest published package to be fetched and executed at install time. If the upstream package is compromised, replaced, or unexpectedly changed, users could run unreviewed code, creating a supply-chain risk even though the README itself is only documentation.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The README says the skill "activates automatically when you need it" and then tells users to "Just describe your situation," but it does not define concrete trigger phrases, scope limits, or non-trigger cases. This is vague enough to overlap with ordinary conversation and could cause unintended invocation.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This script performs repository mutation (git add, git commit, git push) and external publishing, capabilities unrelated to providing ghosting advice. In the context of an end-user skill, such side-effecting operations are more dangerous because they can modify source control state and publish artifacts externally if run by a user who assumes the package only contains conversational logic.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

The script invokes npx clawhub publish without pinning an exact package version, so execution depends on whatever version npx resolves at runtime. This creates a supply-chain risk: a compromised, typosquatted, or unexpectedly changed package version could execute arbitrary code during publishing with the user's local privileges.

Content

No source excerpt is available for this finding.