safety-escalation-review

Process local evidence into redacted safety memos and handoff checklists.

325|60|Updated Apr 20, 2026
One-click install
npx skills add https://github.com/Prompthon-IO/agent-systems-handbook --skill safety-escalation-review
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: safety-escalation-review
Source: https://github.com/Prompthon-IO/agent-systems-handbook/tree/main/skills/safety-escalation-review
Command: npx skills add https://github.com/Prompthon-IO/agent-systems-handbook --skill safety-escalation-review

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Assists teams when evidence suggests credible harm by providing a local-first review pattern, redacted memo generation, and a clear human handoff checklist, without prematurely involving external authorities.

Core Features & Use Cases

  • Local intake and normalization of transcripts, notes, or JSON evidence bundles for offline processing.
  • Deterministic extraction of risk signals and timeline cues, with redaction for privacy.
  • Generation of a Markdown escalation memo and checklist for human reviewers, plus a defined handoff path.
  • Supports human ownership decisions (trust & safety, legal, security, or incident response) while preserving original evidence locally.

Quick Start

Run the escalation reviewer against a local evidence file to generate a redacted safety escalation memo.

Frequently Asked Questions about safety-escalation-review

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prepare a safety escalation memo from local evidence files?

You can generate a redacted escalation memo by processing local transcripts, notes, or JSON evidence bundles to deterministically extract risk signals and timeline cues, producing a markdown document with a human handoff checklist.

What is local evidence redaction for trust and safety reviews?

Local evidence redaction is the offline process of removing sensitive identifiers from transcripts, notes, or JSON bundles while extracting credible harm signals for an escalation memo, ensuring privacy before human handoff.

Can I extract harm risk signals from JSON evidence bundles without external reporting?

Yes, you can extract harm risk signals from JSON evidence bundles entirely offline. The system performs deterministic local processing and redaction to generate a memo, strictly avoiding external reporting or data transmission.

Does this escalation review workflow support incident response handoffs?

Yes, the escalation review workflow supports incident response handoffs by generating a structured human handoff checklist and redacted memo, allowing human reviewers to retain ownership across trust, safety, legal, or security decisions.

What are the limitations of deterministic local processing for safety escalations?

A key limitation is that deterministic local processing strictly avoids external reporting or data transmission. It only prepares a redacted memo and handoff checklist, leaving the actual ownership decision to human reviewers.