Enterprise Multi-Agent Automation Harness

What it does

Reliability patterns for running three or more Claude Code agents unattended: handoff
documents that survive restarts, a task router with a fallback backend, exponential
back-off that refuses to retry authentication errors, a sliding-window rate-limit guard,
JSON-lines audit logging, stale-agent detection with a restart cooldown, and a two-step
tmux notification protocol.

Inputs and outputs

Input: a description or the code of an existing multi-agent pipeline.

Output: a gap table and prioritized fixes, plus copy-ready Python and shell snippets
(retry decorator, rate-limit guard, logger, router, self-healer, health check).

Package contents

• SKILL.md - patterns and full snippets.
• prompt.md - reusable review prompt and output contract.
• tests/test_package.py - package-contract test.
• LICENSE - proprietary distribution notice.

Review and safety disclosure

• No executable files, background services, external endpoints, or credentials are included;
  the snippets are text you copy into your own project.
• The self-healer sends Ctrl-C and a resume command to a tmux pane. Use it only on panes
  you own, and keep the cooldown so it cannot loop.
• Keep secrets in your OS keychain or environment, never in handoff documents or logs.
• The output is an engineering review, not a guarantee of uptime.