name	verify
description	Verify EvalCore end-to-end — format check, clippy, full test suite, and a real CLI run against the quickstart example. Use before declaring any change done, before commits, and after dependency or toolchain changes.

Verify EvalCore

Run every step; do not stop at the first green. Report which steps passed/failed with the failing output.

Format: cargo fmt --all --check
Lint: cargo clippy --workspace --all-targets -- -D warnings
Tests: cargo nextest run --workspace (fallback if nextest missing: cargo test --workspace)
End-to-end smoke (the real binary, no network needed):
```
cargo run -q -p evalcore -- validate examples/quickstart/evals.yaml
cargo run -q -p evalcore -- run examples/quickstart/evals.yaml
```
Expected: validate prints the config summary; run prints per-case PASS lines and a summary with 0 failed, exit code 0.
Exit-code contract (only when engine/CLI code changed): run against a config whose scorer cannot pass (e.g. contains: value: "xyzzy-not-present" in a temp dir) and confirm exit code is exactly 1.

Notes:

Never verify against real LLM APIs; HTTP paths are covered by wiremock tests.
If snapshots fail after an intentional output change: INSTA_UPDATE=always cargo test -p evalcore-report, then re-run normally and include the snapshot diff in your report.

name	verify
description	Verify EvalCore end-to-end — format check, clippy, full test suite, and a real CLI run against the quickstart example. Use before declaring any change done, before commits, and after dependency or toolchain changes.

Run every step; do not stop at the first green. Report which steps passed/failed with the failing output.

Format: cargo fmt --all --check
Lint: cargo clippy --workspace --all-targets -- -D warnings
Tests: cargo nextest run --workspace (fallback if nextest missing: cargo test --workspace)
End-to-end smoke (the real binary, no network needed):
```
cargo run -q -p evalcore -- validate examples/quickstart/evals.yaml
cargo run -q -p evalcore -- run examples/quickstart/evals.yaml
```
Expected: validate prints the config summary; run prints per-case PASS lines and a summary with 0 failed, exit code 0.
Exit-code contract (only when engine/CLI code changed): run against a config whose scorer cannot pass (e.g. contains: value: "xyzzy-not-present" in a temp dir) and confirm exit code is exactly 1.

Notes:

Never verify against real LLM APIs; HTTP paths are covered by wiremock tests.
If snapshots fail after an intentional output change: INSTA_UPDATE=always cargo test -p evalcore-report, then re-run normally and include the snapshot diff in your report.

verify