fix: align ZeroAPI with OpenClaw 2026.4.20

[dorukardahan]

Summary

• switch Kimi starter configs and examples to OpenClaw v2026.4.20's moonshot/kimi-k2.6 default
• keep K2.5 benchmark values as an explicit temporary proxy until the benchmark refresh has native K2.6 data
• document the new cron jobs-state.json split and correct same-provider auth-profile steering docs

Why

OpenClaw v2026.4.20 changed the practical Moonshot default, split cron runtime state out of jobs.json, and still does not merge authProfileOverride from before_model_resolve. ZeroAPI should match those runtime contracts instead of carrying stale K2.5/default and auth-profile wording.

Tests

• npm test
• JSON parse check for policy-families.json, examples/openai-glm-kimi.json, and examples/full-stack.json
• grep check for stale direct-authProfile wording

Non-goals

• no live benchmark refresh in this PR
• no change to OpenClaw provider auth or runtime hook behavior

[dorukardahan]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. More of your lovely PRs please.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".