docs: add Codex and Claude Harness examples and a codex runtime-image conref

[chinmaychahar]

Closes #530

Adds Codex and Claude Harness examples to agents/agent-harness.md and a runtime-image-codex conref that the ref-docs job resolves

Placement: both examples go on the harness page, next to the existing example and the runtime tables, so a reader can compare the runtimes in one place. Open to suggestions.

Done when

• Codex Harness example with its ModelConfig and an AgentTemplate, run on a cluster
• Claude Harness example on the same terms
• runtime-image-codex.md exists and resolves; the other two conrefs still resolve
• runtime-image-claude.md is referenced by a page again
• Every image is digest-pinned
• Placement question answered

Testing

On kind with kagent 1.0.0-alpha9, both Agents reach Ready with no warnings and Sessions start on both. A Codex ModelConfig without apiFormat: responses is rejected

UI

[Rachael-Graham]

I ran both examples on a kind cluster running kagent 1.0.0-alpha8 with Agent Substrate 0.4.0-alpha1, applied exactly as the page renders them.

Codex. All three resources are accepted, the Agent reports Accepted, ResolvedRefs, Compatible and Ready with no warnings, the Session reaches RUNTIME_STATE_READY, and a turn returns the model's reply. The manifest is correct as written.

Claude. Same four conditions, no warnings, and the Session starts and reaches the Anthropic API. I have no Anthropic key, so the turn itself ends at Invalid API key · Fix external API key — everything up to the model call is verified, the model call is not.

Also checked: both manifests pass the alpha8 CRD schemas with no unknown or missing fields and every CEL rule satisfied, the digest you pinned is the one GHCR serves for codex-harness:1.0.0-alpha8, the Secret names and keys match the OpenAI and Anthropic guides, both provider links and the #model-provider-support anchor resolve in a local build, and the page builds clean with both tabs rendering.

The deploy check fails because a pull request from a fork cannot read CLOUDFLARE_API_TOKEN. Nothing for you to fix.

Comments inline.

[chinmaychahar]

@Rachael-Graham thanks for the review! I've addressed them and also rebased onto mainline/alpha9

I re-ran both examples fresh. Updated screenshots

[Rachael-Graham]

Looks good. Running the workflows and once those pass I will merge. Thank you very much!