fix: unset CI/GITHUB_ACTIONS env vars for eval runs

Claude CLI routing behavior changes when CI=true — it skips skill
invocation and uses Bash directly. Unsetting these markers makes
Claude behave like a local environment for consistent eval results.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Garry Tan
2026-03-23 07:38:35 -07:00
parent cb364e1978
commit 56c4066680
+4
View File
@@ -115,6 +115,10 @@ jobs:
OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}
EVALS_CONCURRENCY: "40"
# Unset CI markers so Claude CLI behaves like a local environment
# (CI=true changes Claude's routing/tool selection behavior)
CI: ""
GITHUB_ACTIONS: ""
run: EVALS=1 bun test --retry 2 --concurrent --max-concurrency 40 ${{ matrix.suite.file }}
- name: Upload eval results