Saltar al contenido
TakoAPI
Volver a skills

enterprise-legal-guardrails

por lawyered0

Web & Frontend Development
677Descargas según clawskills.sh
2Estrellas
0Vistas
0 me gustaInicia sesión para dar me gusta a esta skill
—Aún sin valoraciones
Tu valoración

Resumen

Legal/compliance guardrails for outbound OpenClaw actions (anti-spam, defamation, privacy, financial claims).

Descripción

Qué hace esta skill

A preflight compliance checker that screens outbound bot-generated text against legal and policy rules before any post, message, or publish action executes. It classifies draft content as PASS, WATCH, REVIEW, or BLOCK based on configurable profiles covering anti-spam, privacy, HR, market claims, and legal language. Runs as a shared safety layer across any skill that produces outbound content.

Centralizes compliance enforcement in one reusable layer so individual skills don't each need their own policy logic.

Cuándo usarla

  • Blocking a bot from posting a price prediction that could violate financial regulations
  • Flagging a mass-messaging campaign as spam before it sends
  • Catching personally identifiable data in a draft comment before it goes public
  • Routing an HR-related message for human review before delivery
  • Screening market commentary for outcome certainty language before publication

Flujo de trabajo de ejemplo

ENTRADA

User asks: Check this draft market commentary before posting it

AGENTE
  1. 1

    Draft text is passed to the checker with --action market-analysis and --policies market

  2. 2

    Checker evaluates the text against the market and financial profiles

  3. 3

    Checker returns a REVIEW result due to a certainty assertion about price movement

  4. 4

    Agent rewrites the draft to remove the certainty claim

  5. 5

    Checker re-runs and returns PASS, allowing the post action to proceed

SALIDA

Post executes after the revised draft clears the market policy check

Instalación

clawhub install lawyered0/enterprise-legal-guardrails

Ejecuta el comando de la CLI o pega el prompt en Claude Code, Cursor, Codex o el agente de programación que uses.

Enlaces