Most agentic workflows don't crash. They just deliver wrong output

silently — no error, no warning, no trace. By the time you notice,

the damage is three steps downstream.

This automation runs a structured failure audit on any Claude-based

agent workflow. Paste your agent design, system prompt, or workflow

description — and it maps every point where failure can happen without

raising an alert.

What you get:

→ Silent failure inventory: where your agent can fail without telling you

→ Detection gap map: which failure points have zero monitoring

→ Downstream blast radius: what breaks when each silent failure triggers

→ 3 hardening patches: the minimum changes that close the highest-risk gaps

Built for: AI engineers, Claude Code builders, n8n/Make automation

architects, anyone shipping agentic workflows to production.

This is not a generic "test your agent" checklist. It's a structured

diagnostic that distinguishes observable failures from invisible ones —

the difference between what your workflow reports and what it actually does.