fix: harden support wording contract inputs

[Stahl-G]

Summary

• treat explicit unsupported / contradicted / insufficient evidence labels as reader-risk warnings even when CSM policy projection does not mark the atom blocking
• validate Claim Ledger claim entries before support wording projection so malformed claims surface as invalid_claim_ledger instead of checked
• add regressions for non-blocking unsupported labels and invalid Claim Ledger entries through direct/status/Quality Panel surfaces

Boundaries

• projection remains read-only
• no gate execution, delivery approval, release authority, support truth proof, or CSM row generation

Validation

• python3 -m pytest -q tests/test_support_wording.py
• python3 -m pytest -q tests/test_support_wording.py tests/test_quality_panel.py tests/test_status.py tests/test_release_consistency.py
• python3 -m pytest -q
• python3 scripts/check_release_consistency.py --no-tag
• python3 scripts/check_skill_contract.py
• python3 scripts/check_briefloop_skill_freshness.py
• python3 scripts/check_capabilities.py
• python3 -m compileall -q src tests
• git diff --check

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: cc9b8cad85

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".

[chatgpt-codex-connector]

Reject non-string Claim Ledger fields before projecting

When a workspace has a malformed claim_ledger.json where a required field is present but not a string, this new validation still returns checked because ClaimContract.validate() only treats None/blank strings as missing. For example, claim_id: 123 is marked invalid by the artifact registry, but support wording proceeds and later renders the finding with an empty claim_id because _text() drops non-strings, so Status/Quality Panel can present diagnostics from a ledger the control plane considers invalid. Reuse the existing ledger payload validation or add the same required-field type checks before projecting.

Useful? React with 👍 / 👎.