Benchmarks: adopt .evals context naming and the evals CLI

Follow-up to the pr2 rename: instructions, READMEs, and the ale-bench
reference solution now teach `evals run/status/result/submit` and the
.evals/tasks + .evals/results paths; baseline gitignores cover .evals/.
Compiled trees are build outputs (not committed) and pick this up on rebuild.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Author: varunursekar

harness-engineering-bench/ale-bench/environment/Dockerfile

@@ -1,6 +1,6 @@
 # Main service: the optimizer workbench a coding agent works in.
 # Holds the editable target (solution.cpp) plus the VeRO CLI so the agent can
-# call `vero harbor eval` / `vero harbor status` against the trusted sidecar.
+# call `evals run` / `evals status` against the trusted sidecar.
 FROM ghcr.io/astral-sh/uv:python3.12-bookworm
 
 RUN apt-get update \

harness-engineering-bench/ale-bench/instruction.md

@@ -15,11 +15,11 @@ the statement specifies.
 3. Score the current commit on the validation set (public seeds):
 
    ```bash
-   vero harbor eval --backend cmd --evaluation-set ale-bench \
+   evals run --backend cmd --evaluation-set ale-bench \
      --partition validation --start 0 --stop 1
    ```
-   Then `vero harbor eval-status JOB_ID` / `vero harbor eval-result JOB_ID`.
-4. Use `vero harbor status` to see evaluation access. VeRO pools your measurements,
+   Then `evals status JOB_ID` / `evals result JOB_ID`.
+4. Use `evals status` to see evaluation access. VeRO pools your measurements,
    re-scores the best on the held-out private seeds, and ships the best candidate
    (the unmodified baseline is a floor, so a regression is never shipped).
 

harness-engineering-bench/ale-bench/solution/solve.sh

@@ -2,5 +2,5 @@
 set -eu
 cd /work/agent
 git config user.name optimizer; git config user.email o@v.test
-vero harbor eval --backend cmd --evaluation-set ale-bench --partition validation --start 0 --stop 1
-vero harbor status
+evals run --backend cmd --evaluation-set ale-bench --partition validation --start 0 --stop 1
+evals status

harness-engineering-bench/gaia/README.md

@@ -15,10 +15,10 @@ All 165 cases come from the immutable dataset reference recorded in
 available to the optimization agent with full result disclosure. Validation is
 aggregate-only and is used to select candidates. Test is held out until Harbor
 grades the completed outer task. The complete development task packages and
-attachments are mounted read-only under `.vero/cases/`; successful and failed
+attachments are mounted read-only under `.evals/tasks/`; successful and failed
 development evaluations place their complete Harbor trial records—including
 exact failures and target-agent logs—under
-`.vero/evaluations/`. Neither validation nor test resources are mounted.
+`.evals/results/`. Neither validation nor test resources are mounted.
 
 The pinned GAIA tasks declare a 600-second Harbor agent timeout. The build
 config sets `case_timeout_seconds: 180` and

harness-engineering-bench/gaia/baseline/README.md

@@ -34,7 +34,7 @@ program. A separate inference-gateway container holds it. The outer coding
 agent uses a producer-scoped token for `gpt-5.4` or `gpt-5.5`; GAIA candidates use an
 evaluation-scoped token restricted to `gpt-5.4-mini-2026-03-17`. Their request
 and token budgets are recorded separately and are visible through
-`vero harbor status`.
+`evals status`.
 
 Run the outer optimization through VeRO so it can give Harbor's coding-agent
 adapter the scoped credential while forwarding the upstream credential only to
@@ -70,10 +70,10 @@ run it locally against Modal eval backends.
 
 The coding agent edits only `target/`; inner evaluations run the candidate
 against the pinned GAIA tasks on Modal. Complete development tasks and
-attachments are mounted read-only under `.vero/cases/`. After each development
+attachments are mounted read-only under `.evals/tasks/`. After each development
 evaluation, complete Harbor trial records for every case—including exact
 failures and target-agent logs—are available under
-`.vero/evaluations/`. Validation remains aggregate-only, and test remains
+`.evals/results/`. Validation remains aggregate-only, and test remains
 verifier-only.
 
 Before Modal teardown, the shared verifier exports the complete VeRO session

harness-engineering-bench/swe-atlas-qna/baseline/.gitignore

@@ -1,2 +1,3 @@
 compiled/
 .vero/
+.evals/

harness-engineering-bench/tau3/baseline/.gitignore

@@ -1,2 +1,3 @@
 compiled/
 .vero/
+.evals/