Ethical Harness Lite.

Most AI agent failures aren't model failures — they're governance failures. No system prompt scoping what the agent can touch. No audit log when it acts. No escalation protocol when something looks wrong. No charter for what "done" means.

Ethical Harness Lite is the 11-artifact governance toolkit Action AI uses on every internal agent before it ships. Drop the files into a Claude Project, follow the setup guide, and you have a working harness in under an hour.

What's inside the zip:

• Tier 1 · Foundation — HCSI base system prompt, agent charter template, decision & audit log skill
• Tier 2 · Governance — reversibility skill, scoping skill, HCSI memory file
• Tier 3 · Hygiene — approval-gate checklist, escalation protocol, incident report template, quarterly review template
• Reference — failure pattern map, glossary, tier overview one-pager
• Start here — PDF setup guide + compatibility notes (Claude Projects, Skills, Cowork)
• README + LICENSE — install instructions and single-user, no-redistribution terms

Built for: SMB owners, project managers, and AI-curious leaders who are about to deploy their first agent — or just realized the one already running has no harness around it.

Read the setup guide on a Saturday. Install Monday. The kit pairs with AURA Module 01 (the free intro deck) for the full conceptual frame.