Claude Code v2.1.269 adds plugin eval suite, /output-style and Bash edit diffs
Anthropic
Claude Code v2.1.269 (September 11) adds `claude plugin eval` — running a plugin's eval suite against Claude Code with scored reproducible JSON+HTML reports — plus `/output-style [name]` for listing and switching output styles including remote/headless sessions, and opt-in diffs of files changed by Bash commands (`bashEditDiffEnabled`). New settings include `CLAUDE_CODE_WORKFLOW_MAX_CONCURRENT_AGENTS` (1-256) for Workflow parallelism, `OTEL_METRICS_INCLUDE_REPOSITORY`, and a gateway model-discovery timeout override. Fixes cover prompt-cache invalidation after output-token cutoffs, terminal key regressions and `/goal` retry with backoff; VS Code gains an agent map with per-agent cards and stop controls. This follows v2.1.268 (covered yesterday), which fixed the HTTP-400 regression on third-party Anthropic-compatible endpoints.
Why it matters
Plugin evals make agent-extension quality measurable in CI, and the release continues the enterprise-gateway push visible across recent versions.
Importance: 2/5
Minor version with substantive plugin-testing and workflow features