← 2.1.0 Test inconclusive · runtime-test

language setting — runtime test

Hands-on runtime battle-test of language setting. Result: PARTIAL.

Language-setting runtime test found the feature does not work at skill scope, and global scope was left untested.

Test Setup

On 2026-01-07, a skill-level invocation test attempted to enable Japanese-language responses via language: japanese in a test skill's SKILL.md frontmatter. The test created .claude/skills/test-language/SKILL.md with the language field specified, following the changelog claim that language configures Claude's response language.

What Happened

The invoked skill responded in English. The language field in skill frontmatter had no observable effect. This ruled out the hypothesis that language setting works at per-skill granularity.

The test then identified that language is likely a global settings.json field (applied session-wide, not per-task), but did not proceed to validate this with an actual global test. No restart was performed, no settings.json modification was tested.

Findings

Language is not a skill frontmatter field. The changelog references language configuration, yet the skill-scoped mechanism does not work.

The feature's actual scope—global or absent—remains unverified. A complete test would require:

  • Adding "language": "japanese" to ~/.claude/settings.json
  • Restarting Claude Code
  • Verifying response language changed

Implications

If language is truly global-only, per-skill language switching is impossible without workarounds (prompt-based instruction, separate instances). If the global setting also fails, the feature may not be implemented in the version tested. The inconclusive result prevents reliable design guidance for multi-language workflows.

Status: PARTIAL—skill scope tested and failed; global scope identified but untested.

Primary source
⎘ 2.1.0/tests/04-language-setting/TEST-RESULTS.mdverbatim from the corpus

Test Results: Language Setting

Feature: Added language setting to configure Claude's response language (e.g., language: "japanese")

Tested: 2026-01-07

Test Setup

Test 1: Skill Frontmatter

Created .claude/skills/test-language/SKILL.md:

---
name: test-language
description: Test language setting in skill frontmatter
language: japanese
context: fork
---

Result: FAILED - Skill responded in English. Language field in skill frontmatter not recognized.

Test 2: Global Settings

Checked ~/.claude/settings.json for language field - not currently set.

Findings

  1. Language is NOT a skill frontmatter field

    • Adding language: japanese to skill had no effect
    • Skill responded in English
  2. Language is likely a global settings.json field

    • Per changelog: "language setting to configure Claude's response language"
    • Would be set at ~/.claude/settings.json level
    • Affects all responses, not per-skill
  3. Not tested: global language setting

    • Would need to add "language": "japanese" to settings.json
    • Then restart Claude Code
    • Verify responses are in Japanese

Implications

  • Language is session-wide, not task-specific
  • Cannot have per-skill language switching (as tested)
  • For multi-language workflows, would need workaround:
    • Prompt-based language instruction
    • Separate Claude instances with different settings

Status: PARTIAL - NEEDS GLOBAL TEST

Skill-level language doesn't work. Global setting not tested (would affect all responses).

Evidence & receipt
◇ ed25519 receipt
idtest_70ba39fc759767761f6e9098
alged25519
pubkey9b87705613b1e2fd064d57fa75a6b679d2856ceafad6b1daa8f982493871b6dd
sig88acd81e055d5f6b895a8b368881be77f9bfe84bfa060e9b678fbf4dbe74ad53a30b4fd057aa364b83a5ee96779682262833f74b913663701357e4d695dd0f04

Signed with an ed25519 key held off the repo. Anyone can verify against the published public key; nobody without the secret key can forge it. Click verify: it recomputes the signature in your browser. The signature proves integrity and authorship of this exact content — not a third-party timestamp or that the underlying claim is objectively true. signedAt is when the @f3/attest pipeline ran, not when the work happened; the evidence refs carry the source dates.

Connected