Skip to content

Retire the tests/manual/ manual-smoke layer — automated L0–L6 is the sole gate #892

@ZaxShen

Description

@ZaxShen

Human-directed (2026-06-18): retire the manual-smoke test layer. Like the arch-docs design (#135), it's a stale layer — the automated L0–L6 (Docker install-smoke, lint, MCP unit+integration, workflow-sim, L5 per-flow eval, L6 multi-turn real-CC, release canary) + the world model now cover what the human-walk did. No "manual" anything remains in the project.

Scope

  • DELETE the tests/manual/ directory entirely (scenarios.md, README.md, setup.md, mcp-health-hook.md, mcp-readonly-fallback.md, task-batch-size-instrumentation.md, debug-mode-expand.md).
  • Strip references to the manual-smoke LAYER (not unrelated English uses of the word "manual"):
    • CONTRIBUTING.md (~73 the RC manual spot-check, ~88 "walk scenarios.md before PR") → reframe: the CI release-gate (L0–L6) is THE gate; drop the manual spot-check.
    • tests/README.md (~21 layer-table Manual row, ~69 tree, ~181, ~247) → remove the Manual-smoke layer from the testing-philosophy/layer list.
    • tests/run-all.sh (~14 the "Manual smoke (fallback)" comment line).
    • tests/l1-lint/no-developer-paths.sh (~7 comment + ~85 the tests/manual allowlist entry).
  • Do NOT touch unrelated "manual" words (e.g. "manual nuke", "manually", a manual-test-A session_id in other fixtures) — only the manual-smoke layer + its dir.

Acceptance

  • tests/manual/ deleted; no reference to the manual-smoke layer remains; testing philosophy in CONTRIBUTING/tests-README reads as "automated L0–L6 is the gate" with no manual layer.
  • L0–L4 green (the deleted dir isn't referenced by any runner; run-all.sh + no-developer-paths.sh still pass).
  • Sequence AFTER ✂️ chore(claude.md): extract trajectory DB definition to top-of-file glossary #135 task 228 merges (228 also touches tests/EVALUATION + tests/ — avoid the overlap).

Milestone v0.10.0. Sibling of #135 (stale-design retirement).

Metadata

Metadata

Assignees

No one assigned

    Labels

    DoctrineDoctrine clarification or contract changeImprovementRefactor / polish / quality workTestsTest infrastructure (L0-L6)

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions