Pular para o conteúdo
TakoAPI
Voltar às skills

clawdstrike-test

por misirov

Security & Passwords
1.3kDownloads informados pelo clawskills.sh
0Visualizações
0 curtidaEntre para curtir esta skill
—Ainda sem avaliações
Sua avaliação

Resumo

Security audit and threat model for OpenClaw gateway hosts.

Descrição

O que esta skill faz

Clawstrike audits an OpenClaw deployment for misconfigurations and exploitable attack paths. It collects system evidence through a verified script, then evaluates gateway exposure, filesystem hygiene, plugin supply chain, and network posture. The output is a structured OK/VULNERABLE report with severity ratings, evidence excerpts, and remediation steps.

It enforces a strict verified-collection mode and redaction rules, which prevents the audit itself from becoming an attack surface or leaking secrets during the review.

Quando usá-la

  • Checking an OpenClaw host before exposing it to a new network
  • Reviewing third-party skill/plugin files for supply chain risks
  • Verifying firewall and port configuration on a self-hosted gateway
  • Auditing filesystem permissions and SUID/SGID binaries on a gateway host
  • Generating a security report for a team handoff or compliance review

Exemplo de fluxo de trabalho

ENTRADA

User asks: run a full security audit of my OpenClaw gateway

AGENTE
  1. 1

    Executes scripts/collect_verified.sh in the working directory to gather evidence

  2. 2

    Reads verified-bundle.json and builds the report header with timestamp, OS, version, and config path

  3. 3

    Evaluates every check in references/required-checks.md against collected evidence

  4. 4

    Applies threat model from references/threat-model.md aligned with findings

  5. 5

    Emits a findings table with severity, evidence excerpts, and fix instructions

SAÍDA

A structured OK/VULNERABLE report listing each check result with redacted evidence and specific remediation steps for any failures

Instalação

clawhub install misirov/clawdstrike-test

Execute o comando da CLI ou cole o prompt no Claude Code, Cursor, Codex ou no agente de código que você usa.

Links