Skip to content

Ghost Security Skills - Setup Guide

Source: ghostsecurity/skills via skills.sh - 28.0K combined installs across 8 indexed listings; first seen Oct 9, 2026 (extended-query probe) GitHub: ghostsecurity/skills (409 stars, Apache-2.0 license; pushed Sep 28, 2026; plugins/<plugin>/skills/<name>/SKILL.md layout) Category: Application Security / DevSecOps Quality Tier: 🟡 Beta - company-backed (Ghost Security); Apache-2.0; mixed sampled verdicts (Socket Warn) - see Security

Ghost Security, an AppSec vendor, publishes this Claude Code plugin marketplace with two plugins: ghost (the AppSec skill set) and exo (workflow orchestration on the vendor's platform). The AppSec plugin wraps four deterministic engines in an AI skills layer: Poltergeist (secret scanner with dual-engine pattern matching and entropy analysis), Wraith (dependency scanner powered by the OSV database), Reaper (MITM HTTPS proxy for live vulnerability validation), and Exorcist (AI-powered code analysis covering 102 vulnerability types). The flagship listing, ghost-scan-code, sits at 4,860 installs, and all eight indexed listings are above 685.

The architecture the README stresses: real tools produce real data, and AI adds judgment on top. Pattern matches, CVE lookups, and traffic captures stay deterministic and auditable; the skills then assess exploitability and return findings with remediation guidance the agent can apply. The skills follow a find, validate, fix loop, results are cached on your machine, and the underlying tools can also be used standalone.

Installation

Claude Code plugin marketplace:

claude plugin marketplace add ghostsecurity/skills
claude plugin install ghost@ghost-security
claude plugin install exo@ghost-security

Or install from inside a Claude Code session:

/plugin marketplace add ghostsecurity/skills
/plugin install ghost@ghost-security
/plugin install exo@ghost-security

Install only the plugins you need. If the install summary asks for it, run /reload-plugins to activate the plugin. The README's installation doc covers loading plugins without installing them, where data is stored, and how to update.

What It Provides

Install counts from the Oct 10, 2026 skills.sh snapshot:

Skill Installs Use For
ghost-scan-code 4,860 AI-powered detection of code security issues (SAST)
ghost-scan-secrets 4,144 Context assessment of detected secrets and credentials
ghost-scan-deps 4,023 Exploitability analysis of dependency vulnerabilities (SCA)
ghost-proxy 3,666 HTTP proxy that powers the ghost-validate live-validation skill
ghost-validate 3,567 Dynamic validation of findings against a live application (DAST)
ghost-report 3,559 Combined security report across all scan results
ghost-repo-context 3,542 Build shared repository context: business criticality, sensitive data, and component map
ghost-exo 685 Build, improve, and debug workflows on the exo agent orchestration platform

All 8 indexed listings clear the 500-install threshold and appear in the table above.

Why This Matters for Hermes Agents

Agent-written code ships faster than humans can audit it, and this plugin targets exactly that gap with an agent-native AppSec pipeline. The split architecture is the interesting part: the scanners are deterministic binaries whose output is auditable, while the AI layer only adds context and exploitability judgment, so findings arrive with a reason and a fix path instead of raw alerts. The skills map onto how coding agents already work: ghost-repo-context to learn the codebase, ghost-scan-secrets and ghost-scan-deps for fast wins, ghost-scan-code for SAST depth, ghost-validate to prove a finding against a running app, then ghost-report to summarize. Rules and criteria are YAML files you can extend or replace, and the tools work standalone, so one scanner can be adopted without the full pipeline. For Hermes agents, the pattern generalizes: deterministic tools plus an orchestration prompt is the same shape as a well-built multi-tool agent workflow.

Usage

You say What happens
"Scan this repo for leaked secrets before I push" ghost-scan-secrets runs Poltergeist and assesses each match in code context
"Are any of our dependencies exploitable?" ghost-scan-deps checks lockfiles through Wraith and OSV, then ranks exploitability
"Give this service a security review" ghost-scan-code runs AI-powered SAST detection over the codebase
"Prove this finding is real against the running app" ghost-validate uses Reaper's proxy to dynamically validate findings (DAST)
"Summarize this sprint's security findings" ghost-report combines results from all engines into one report
"What should we threat-model first in this repo?" ghost-repo-context maps business criticality, sensitive data, and components
"Build a workflow on exo for triaging scans" ghost-exo routes the request to build, improve, or debug intents on the exo platform

Verification

After installation, open Claude Code and confirm ghost and exo appear as installed plugins, running /reload-plugins if the install summary asks for it. Review the skill source before installing:

# Fetch the layout-verified scan-code SKILL.md straight from GitHub
curl -sL https://raw.githubusercontent.com/ghostsecurity/skills/main/plugins/ghost/skills/scan-code/SKILL.md | head -20

Security

skills.sh verdicts for sampled skills (verified Oct 10, 2026) - re-check the security pages on skills.sh before production use:

Skill Gen Agent Trust Hub Socket Snyk
ghost-scan-code Pass Warn Pass
ghost-scan-deps Pass Warn Warn
ghost-scan-secrets Pass Pass Pass

Two of the three sampled skills carry Socket Warns, and ghost-scan-deps also scores a Snyk Warn; the sample covers three of the eight listings, so extend the review before wiring the plugin into CI.

Integration with CorpusIQ

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects

Limitations

  • Mixed sampled verdicts: Socket Warn on ghost-scan-code and ghost-scan-deps; Snyk Warn on ghost-scan-deps. The sample is three skills.
  • The primary flow is the Claude Code plugin marketplace; this is a claude plugin install, not a generic SKILL.md drop-in.
  • Wraith queries the OSV database over the network and the scanner engines ship as GitHub release binaries; air-gapped setups need extra planning.
  • Young project from a single vendor: 409 stars, first indexed Oct 9, 2026; the exo plugin ties into the vendor's own orchestration platform.
  • Snapshot data, verified Oct 10, 2026: 28,046 combined installs across 8 indexed listings; 409 GitHub stars; Apache-2.0; last pushed Sep 28, 2026. Counts drift over time.

Pro Tips

  • Run ghost-repo-context first so scans can weigh findings against business criticality and sensitive-data locations.
  • Keep ghost-validate for systems you are authorized to test; it exercises a live application through the proxy.
  • Start with ghost-scan-secrets standalone; Poltergeist works without the rest of the pipeline.
  • Skip the exo plugin unless you use Ghost Security's orchestration platform; the AppSec skills install independently.
  • Re-check the skills.sh security pages before production use: two of the three sampled skills scored Socket Warn.

← Skills Catalog | Skills Marketplace → Powered by CorpusIQ

AI answers on your data. See where every number comes from.

Connect your tools and ask in ChatGPT, Claude, or Perplexity. Enterprise-grade answers in minutes. From $29.95/month.

Start free, no credit card
Contact Terms Privacy Pricing