feat(e2e): add workload catalog

[PsiACE]

Stack

1. #1229 RFC: long-horizon workloads
2. #1230 Bub event capture
3. #1231 Workload catalog
4. #1232 Long-horizon Bub runtime

Merge in this order. This PR depends on #1230.

Which issue or RFC does this PR close?

Implements the workload catalog layer of #1229.

Rationale for this change

Workload discovery should be data-driven and independent of Harbor and runner internals.

What changes are included in this PR?

• Add a pure workload catalog.
• Add manifests for LOCOMO, project decision, and deterministic database recovery workloads.
• Add local Harbor task fixtures for deterministic acceptance.
• Add catalog behavior coverage.

Are there any user-facing changes?

No existing CLI or CI behavior changes in this layer.

How was this change tested?

• make harness-check

AI usage statement

OpenAI Codex (GPT-5) assisted with implementation and tests. The author is responsible for the submitted changes.

[Copilot]

Pull request overview

Introduces a validated, data-driven “workload catalog” layer for the Bub e2e harness, aligning with the long-horizon workloads RFC by defining a portable manifest contract plus deterministic local fixtures.

Changes:

• Added a Pydantic-backed catalog module to load and select workload manifests by ID/category.
• Added three workload manifests (two deterministic acceptance samples, one model-backed long-horizon Terminal-Bench task).
• Added local Harbor task fixtures for deterministic workloads, plus an initial catalog behavior test.

Reviewed changes

Copilot reviewed 17 out of 17 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
e2e/bub/src/powercontext_e2e/catalog.py	Defines the workload manifest schema and implements load/select behavior.
e2e/bub/tests/test_workload_catalog.py	Adds initial coverage for manifest loading and selection behavior.
e2e/bub/tasks/locomo-support-group.yaml	Adds an acceptance/sample workload manifest backed by a local Harbor task fixture.
e2e/bub/tasks/project-database-decision.yaml	Adds an acceptance/sample/smoke workload manifest backed by a local Harbor task fixture.
e2e/bub/tasks/terminal-bench-db-wal-recovery.yaml	Adds a model-backed long-horizon workload manifest using the Terminal-Bench dataset.
e2e/bub/harbor-tasks/locomo-support-group/task.toml	Defines the Harbor task steps for the LoCoMo acceptance fixture.
e2e/bub/harbor-tasks/locomo-support-group/environment/Dockerfile	Provides the container environment for the LoCoMo fixture task.
e2e/bub/harbor-tasks/locomo-support-group/steps/capture/instruction.md	Capture-step instruction for the LoCoMo fixture (PowerContext remember tool call).
e2e/bub/harbor-tasks/locomo-support-group/steps/capture/tests/test.sh	Deterministic verifier output for the LoCoMo capture step.
e2e/bub/harbor-tasks/locomo-support-group/steps/recall/instruction.md	Recall-step instruction for the LoCoMo fixture (PowerContext context tool call).
e2e/bub/harbor-tasks/locomo-support-group/steps/recall/tests/test.sh	Deterministic verifier output for the LoCoMo recall step.
e2e/bub/harbor-tasks/project-database-decision/task.toml	Defines the Harbor task steps for the project decision acceptance fixture.
e2e/bub/harbor-tasks/project-database-decision/environment/Dockerfile	Provides the container environment for the project decision fixture task.
e2e/bub/harbor-tasks/project-database-decision/steps/capture/instruction.md	Capture-step instruction for the project decision fixture.
e2e/bub/harbor-tasks/project-database-decision/steps/capture/tests/test.sh	Deterministic verifier output for the project decision capture step.
e2e/bub/harbor-tasks/project-database-decision/steps/recall/instruction.md	Recall-step instruction for the project decision fixture.
e2e/bub/harbor-tasks/project-database-decision/steps/recall/tests/test.sh	Deterministic verifier output for the project decision recall step.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Pull request overview

Copilot reviewed 17 out of 17 changed files in this pull request and generated no new comments.

Suppressed comments (2)

e2e/bub/tests/test_workload_catalog.py:28

• This test covers the happy-path selection, but it doesn't assert the documented default behavior (no selectors => acceptance) or that unknown IDs/categories raise a clear boundary error. Adding these checks would better protect the catalog contract from regressions.

    acceptance = select_tasks(tasks, categories=("acceptance",))

    assert [task.id for task in selected_ids] == [
        "locomo-support-group",
        "terminal-bench-db-wal-recovery",

e2e/bub/src/powercontext_e2e/catalog.py:104

• When loading a directory of manifests, any YAML/Pydantic failure currently raises without indicating which file caused the error, which makes diagnosing catalog issues harder as the task set grows. Consider validating per-file and wrapping exceptions with the manifest path.

def load_tasks(path: Path) -> tuple[E2ETask, ...]:
    task_paths = sorted(path.glob("*.yaml")) if path.is_dir() else [path]
    tasks = tuple(E2ETask.model_validate(yaml.safe_load(item.read_text(encoding="utf-8"))) for item in task_paths)
    ids = [task.id for task in tasks]
    if not tasks:
        raise ValueError(f"No e2e workload manifests found at {path}")  # noqa: TRY003