fix: reconcile the optimize-skill branch with main's CE035 and run-limit work

Integration fixes the rebase onto main exposed, in one place rather than
smeared across 175 replayed commits.

**The CE035 collision.** Main shipped CE035 as the workflow-output resolver
(`tests/lint/workflow_outputs.py`) while this branch shipped CE035 as the
split-label rule. A `# noqa` keys on the id, so one suppression would disarm
both — exactly what `test_rule_ids_are_unique_across_baserules_and_test_classes`
says. The unmerged one renumbers: the split-label rule is **CE060** now, in the
test class, CLAUDE.md, `tasks/skills/ci-outcome.yaml`, tutorial 09 and
`plan_command`'s pointer. Main's CE035 keeps its number and its file.

Main's `.claude/harness-candidates.md` also reserved four numbers this branch has
since implemented (CE036, CE037, CE040, CE041) plus CE042 for the sleep-loop
candidate. Those reservations move to CE061–CE065, because the backlog is one of
the two places CLAUDE.md tells an author to grep before claiming a number, and a
reservation pointing at a shipped rule is worse than no reservation.

**Main's new code under this branch's new rules.** `task_loader`'s
`system_prompt_file` seam (main, #92) is a `model_copy(update={...})`, which CE048
now forbids outside `models/copy_with.py`; it becomes `copy_with(...)`, whose
delegation to `model_copy` keeps the validator-skipping the two comments there
describe — so the comments were corrected rather than left half-true.

**Two conflicts resolved toward keeping both behaviours**, noted here because the
replayed commits do not say so: `criteria/agent_judge.py` forces main's
`system_prompt_file=None` / `system_prompt_mode="replace"` in the SAME
`model_validate` payload as the branch's CE048 conversion (they are mutually
exclusive under `validate_assignment=True`, so a sequential assignment raises),
and `plan`'s per-variant `validate_run_limits` warnings (main) moved into the
branch's `_report_variants`, escaped for CE050.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: uipreliga

.claude/harness-candidates.md

@@ -384,8 +384,9 @@ with the two `action.yml` items above — one considered change to the action's
   `timeout is None`. Caught independently by two reviewers (`bai-uipath`, `uipreliga`)
   on the PR, both citing the exact same arithmetic mismatch. **Not promoted in this
   pass**, but a stronger candidate than most entries here: `uipreliga` proposed a
-  generic whole-tree rule (proposed as CE035, renumbered CE042 here — CE035 shipped as
-  the workflow-outputs resolver on the published-action branch) — for every sleep-loop under
+  generic whole-tree rule (proposed as CE035, renumbered CE065 here — CE035 shipped as
+  the workflow-outputs resolver on the published-action branch, and CE042 as the
+  replicate-padding seam on the optimize-skill branch) — for every sleep-loop under
   `src/coder_eval/agents/**`, assert its own cycle-count × interval either references a
   timeout-derived name or is provably below `experiments/default.yaml`'s baseline — that
   would catch this class of bug in ANY agent, not just this one (confirmed zero
@@ -395,7 +396,7 @@ with the two `action.yml` items above — one considered change to the action's
 
 ## From 2026-08-04 published-action verification review
 
-- [ ] **CE041 — `VAR=$(… | grep …)` under `set -e` followed by an emptiness check
+- [ ] **CE061 — `VAR=$(… | grep …)` under `set -e` followed by an emptiness check
   is a dead diagnostic.** With `set -euo pipefail`, a pipeline whose `grep` matches
   nothing exits 1, so the assignment aborts the step *before* the
   `if [ -z "$VAR" ]; then echo "::error::…"` branch that was written to report it —
@@ -407,22 +408,22 @@ with the two `action.yml` items above — one considered change to the action's
   `verify-published-action.yml`; **`actionlint` + shellcheck do NOT flag it**
   (verified against the exact snippet), so the actionlint candidate above does not
   subsume this one.
-- [ ] **CE036 — ban the skipped-green job gate.** Fail a job-level `if:` in
+- [ ] **CE062 — ban the skipped-green job gate.** Fail a job-level `if:` in
   `.github/workflows/**` whose only discriminator is an emptiness/equality test on
   `needs.<job>.outputs.<key>`. A lost output on a partial "Re-run failed jobs" resolves
   the job to SKIPPED-**green**, so an operator sees a green re-run while nothing ran.
   Fixed by hand twice now: `promote` was designed around the hazard, and
   `publish-pypi`'s `if: needs.release.outputs.version != ''` (dead *and* dangerous — a
   skipped publish also skipped `promote`) was removed in the follow-up review. CE035
   catches the *typo* class; this catches the *shape*. Escape hatch: inline
-  `# noqa: CE036 — <reason>` for value-driven gates that cannot strand a release.
-- [ ] **CE037 — `if: failure()` is wrong in a job containing a `continue-on-error`
+  `# noqa: CE062 — <reason>` for value-driven gates that cannot strand a release.
+- [ ] **CE063 — `if: failure()` is wrong in a job containing a `continue-on-error`
   step.** Require `always()` (or a reference to the tolerated step's
   `steps.<id>.outcome`) on diagnostic/upload steps in such a job. Fixed by hand in
   `verify-published-action.yml`: the run dir was discarded in exactly the tolerated-red
   case the gate is designed around, because a tolerated red leaves the job green and
   `failure()` never fires. Pure YAML shape check, ~30 lines.
-- [ ] **CE040 — cap inline `run:` bodies; oversized decision logic belongs in
+- [ ] **CE064 — cap inline `run:` bodies; oversized decision logic belongs in
   `.github/scripts/`.** `verify-published-action.yml`'s parity step (~70 lines, 7
   decision points) and its e2e gate (~66 lines, switching from bash to a `python3`
   heredoc mid-step) are 10-20-branch units invisible to `make check`, `make lint`,

CLAUDE.md

@@ -58,7 +58,7 @@ dataset:
 
 Label every row or none. A *partly* labelled dataset is the one genuinely bad state: `--split`
 keeps the rows that match and silently drops the unlabelled ones, so the run succeeds and every
-metric is computed over a smaller suite than the file suggests. `CE035` fails the build on it.
+metric is computed over a smaller suite than the file suggests. `CE060` fails the build on it.
 
 ### One fixture, and the variation lives in the prompt
 

docs/tutorials/09-optimizing-a-skill-body.md

@@ -90,7 +90,7 @@ def _preview_dataset(
     # already in hand.
     rows = load_dataset_rows(task.dataset, task_file.parent)
     # `row_split_label` is the runtime's single definition of "labelled" (select_rows and
-    # CE035 call it too) — do not re-derive the rule here.
+    # CE060 call it too) — do not re-derive the rule here.
     labels = [row_split_label(r, task.dataset.split_field) for r in rows]
     labelled = [x for x in labels if x is not None]
 

src/coder_eval/cli/plan_command.py

@@ -21,6 +21,7 @@
     TaskDefinition,
     TemplateDirSource,
     TemplateSource,
+    copy_with,
 )
 
 
@@ -289,10 +290,10 @@ def resolve_agent_system_prompt[T: AgentConfig | BaseAgentConfig | None](agent_c
     if not prompt_path.exists():
         raise FileNotFoundError(f"system_prompt_file not found: {prompt_path}")
     # A whitespace-only file is no prompt at all — mirror the normalization
-    # _blank_prompt_is_no_prompt applies to inline prompts (model_copy skips
-    # validators, so this seam has to apply it itself).
+    # _blank_prompt_is_no_prompt applies to inline prompts (copy_with delegates to
+    # model_copy, which skips validators, so this seam has to apply it itself).
     content = prompt_path.read_text(encoding="utf-8").strip() or None
-    # ...which means model_copy also skips check_replace_mode_has_prompt, so a
+    # ...which means the copy also skips check_replace_mode_has_prompt, so a
     # blank file under `replace` would reach the agent as (replace, no prompt)
     # and silently downgrade to the append preset at runtime. Reject it here
     # instead: the file is the only prompt the config had, and the docs promise
@@ -302,7 +303,7 @@ def resolve_agent_system_prompt[T: AgentConfig | BaseAgentConfig | None](agent_c
             f"system_prompt_file {prompt_path} is empty; system_prompt_mode='replace' requires a "
             + "prompt to replace the Claude Code default with"
         )
-    return agent_config.model_copy(update={"system_prompt": content, "system_prompt_file": None})
+    return copy_with(agent_config, system_prompt=content, system_prompt_file=None)
 
 
 def resolve_system_prompt_files(task: TaskDefinition, base_dir: Path) -> TaskDefinition:

src/coder_eval/orchestration/task_loader.py

@@ -111,7 +111,7 @@ run_limits:
 # instruction in the body observable.
 #
 # The fixture lives under `templates/`, NOT beside this file, and that is deliberate:
-# anything matching `tasks/**/*.yaml` is loaded as a real TaskDefinition by CE034, CE035,
+# anything matching `tasks/**/*.yaml` is loaded as a real TaskDefinition by CE034, CE060,
 # test_tags and test_yaml_migration — and this fixture must contain an experiment file,
 # which is not a valid task. A fixture placed beside its suite would break all four.
 sandbox:

tasks/skills/ci-outcome.yaml

@@ -3027,7 +3027,7 @@ def test_outcome_rows_are_all_positive(self):
         assert rows, "the outcome template ships no rows"
         # `row_split_label`, not `r.get("split")`: truthiness would report a legitimate
         # `"split": 0` as unlabelled, which is a SECOND definition of "labelled" competing
-        # with the runtime's. CE035 exists precisely to have one.
+        # with the runtime's. CE060 exists precisely to have one.
         assert all(row_split_label(r, "split") is not None for r in rows), (
             "every template row must carry a `split` — a PARTLY labelled dataset is the one bad "
             "state: --split keeps the matching rows and drops the unlabelled ones, shrinking the "
@@ -3278,7 +3278,7 @@ def test_activation_rows_split_both_polarities_both_sides(self):
         ]
         # `row_split_label`, not `r.get("split")`: truthiness would report a legitimate
         # `"split": 0` as unlabelled, which is a SECOND definition of "labelled" competing
-        # with the runtime's. CE035 exists precisely to have one.
+        # with the runtime's. CE060 exists precisely to have one.
         assert all(row_split_label(r, "split") is not None for r in rows), (
             "every template row must carry a `split` — a PARTLY labelled dataset is the one "
             "bad state: --split keeps the matching rows and drops the unlabelled ones, "
@@ -6469,8 +6469,8 @@ def test_finding_reports_the_line_of_the_offending_reference(self, tmp_path: Pat
 
 
 @pytest.mark.lint
-class TestCE035SplitLabelsAllOrNothing:
-    """CE035 — a dataset's split field must be on every row or on none, never on some.
+class TestCE060SplitLabelsAllOrNothing:
+    """CE060 — a dataset's split field must be on every row or on none, never on some.
 
     `optimize-skill` calls a partly-labelled dataset "the dangerous state, because it does
     not look like one", and it is right: ``--split`` keeps the rows whose label matches and
@@ -6765,9 +6765,9 @@ def test_ce036_exemption_list_matches_claude_md(self):
 
 
 def _dataset_task(rows: list[dict], *, prompt: str = "${row.id}", criteria=None, split_field: str = "split"):
-    """A minimal dataset-backed task over inline rows, for the CE035/CE036 fixtures.
+    """A minimal dataset-backed task over inline rows, for the CE060/CE036 fixtures.
 
-    One builder for both rules: CE035 needs varying rows, CE036 varying prompts AND
+    One builder for both rules: CE060 needs varying rows, CE036 varying prompts AND
     criteria, and a per-class copy would fork the moment either grew a parameter.
     Inline ``rows`` deliberately — no JSONL file, so the fixtures never touch disk.
     """
@@ -8940,7 +8940,7 @@ class TestCE052TemplateTasksLoad:
     """CE052 — every task YAML under `templates/` must load through the real `load_task`.
 
     A `@pytest.mark.lint` class rather than a `BaseRule`: it reasons over YAML trees and needs the
-    loader itself, not one `.py` AST at a time — the same shape as CE035/CE036.
+    loader itself, not one `.py` AST at a time — the same shape as CE060/CE036.
 
     **What it caught.** `templates/ci-outcome-fixture/evals/activation.yaml` declared
     `suite_thresholds: {recall.yes: 0.7}` with no `dataset:` block, which