Cross-Agent Model Onboarding Pack 20260818 — Release Notes

What it does

Cross-Agent Model Onboarding Pack gives teams a consistent readiness check for onboarding models across multiple agent platforms. It turns redacted notes into evidence covering platform, provider, model identity, authentication references, capability tests, tool-call behavior, limits, cost ceilings, fallback routing, failure evidence, and rollback.

The skill is useful when model access spans more than one runtime or operator group. It provides a practical advisory verdict without calling providers, reading credentials, or changing configuration.

Use cases

• Platform teams onboarding a new model route across OpenClaw, Hermes, or other agent runtimes.
• Operations teams documenting fallback, rate-limit, context-limit, and rollback evidence before cutover.
• Governance reviewers checking that model capability and tool-call tests are backed by evidence.

Requirements

• Python 3 with standard-library modules only.
• Redacted onboarding notes supplied through `--text` or a single local `--notes` file.
• No API keys required.
• OpenClaw workspace capable of running local skill scripts.

Example usage

python3 scripts/cross_agent_model_onboarding_pack.py --text "agent platform documented; provider identifier recorded; model identifier recorded; authentication reference redacted; capability test passed; tool call test passed; context limit documented; rate limit documented; cost ceiling approved; fallback route tested; failure evidence captured; rollback plan ready" --json

Expected output: JSON with a readiness verdict, matched onboarding evidence signals, missing items if any, and `mutated_files: []`.

Price

$4.99