docs: make CLAUDE.md a compatibility shim

[jbeckwith-oai]

Summary

• replace bare CLAUDE.md pointer with an explicit compatibility shim to AGENTS.md

Test Plan

• not run (docs-only change)

[mldangelo-oai]

@AGENTS.md should be sufficient here

[chatgpt-codex-connector]

💡 Codex Review

promptfoo-python/.github/workflows/test.yml

Lines 305 to 309 in f782023

	if [[ "$LINT_RESULT" == "failure" || "$LINT_RESULT" == "cancelled" ||
	"$TYPE_CHECK_RESULT" == "failure" || "$TYPE_CHECK_RESULT" == "cancelled" ||
	"$TEST_RESULT" == "failure" || "$TEST_RESULT" == "cancelled" ||
	"$TEST_NPX_FALLBACK_RESULT" == "failure" || "$TEST_NPX_FALLBACK_RESULT" == "cancelled" ||
	"$BUILD_RESULT" == "failure" || "$BUILD_RESULT" == "cancelled" ]]; then

Fail CI when classify-changes job fails

ci-success now depends on classify-changes, but its gate only checks lint/type/test/build results. If classify-changes errors (e.g., bad SHA diff), downstream jobs are skipped and this step still exits 0, producing a false green required check. Include needs.classify-changes.result in the failure condition.

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".