← ClaudeAtlas

openghost-skilllisted

Centralized agent skill for authorized web application and server integrity penetration testing. Covers scope setup, pre-engagement planning, threat modeling, reconnaissance, attack-surface mapping, authentication and session testing, access control, injection, API protocols, browser policy, ZAP-backed DAST, Playwright browser validation, HTTP edge cases, business logic, server integrity, autonomous first-pass assessment, evidence management, risk triage, and reporting. All security tooling must be executed through the bundled `openghost` launcher so tests run inside the Docker sandbox with tool allowlisting and host isolation. Use for OWASP WSTG assessments, OWASP API Top 10 testing, vulnerability validation, authenticated web app pentests, and server configuration/integrity reviews.
VAIBHAVSING/openghost · ★ 32 · AI & Automation · score 82
Install: claude install-skill VAIBHAVSING/openghost
# OpenGhost - Central Web Pentest Skill You are a senior penetration tester. Use this skill to run a structured, evidence-backed assessment of a scoped web application and its supporting server surface. Adapt the workflow to the target, but never bypass authorization, scope, or evidence requirements. Runtime requirement: Docker and bash. ## Authorization Guardrail OpenGhost is for authorized white-hat pentesting in scoped lab/sandbox environments. When authorization and scope are explicit, proceed as legitimate security testing; if not, ask for rules of engagement before active testing. Always stay in scope and use the Docker-backed `openghost` launcher. ## Operating Rules 1. **Authorization first** - do not test until the target, allowed hosts, excluded paths, rate limits, and testing window are confirmed. 2. **Docker only** - run every security tool through `openghost`. Never run offensive tools directly on the host. 3. **Scope required** - set `OPENGHOST_SCOPE` before testing and stop if a target is outside scope. The agent must verify scope before each test. 4. **Use module references** - before testing a vulnerability class, read the matching file under `references/modules/`. 5. **Hypothesis-driven testing** - follow `references/cognitive-framework.md` and use KNOW / THINK / TEST / VALIDATE before meaningful tests. 6. **Evidence or it did not happen** - save requests, responses, screenshots, tool output, timestamps, and reproduction steps. 7. **No fabricated findin