Tasks: Implement-Loop Friction Quick-Wins II

Mission: loop-friction-quickwins-2-01KXBWA4 | Branch: feat/loop-friction-quickwins-2 Plan: plan.md (7-concern map, post-squad) | Spec: spec.md

7 work packages, one per Implementation Concern. Each is independently implementable and file-disjoint (no owned_files overlap). WP07 depends on WP01 (reuses its canonical runtime-field helper). Every WP carries a red-first regression proving the guard's true-positive still fires (NFR-005 / C-001).

Subtask Index

IDDescriptionWPParallel
T001Canonical runtime-field accessor from frontmatter.py::WP_FIELD_ORDERWP01
T002_drop_runtime_frontmatter_only_wp + wire into resolve_planning_artifact_stagingWP01
T003Test: runtime-only WP diff dropped; N-lane 0-commit allocationWP01
T004Test: body/non-runtime change still blocks; no-op when auto_commit=TrueWP01
T005Extend _normalize_tasks_md to canonicalize pipe-table status cellsWP02[P]
T006Test: pipe-table [D]/[P] + mixed churn stays currentWP02[P]
T007Test: substantive tasks.md row change still stales (true-positive)WP02[P]
T008Interpreter-resolution helper (uv run → sys.executable fallback)WP03[P]
T009Apply helper + contention lock (own timeout, decoupled from run timeout)WP03[P]
T010Test: pytest-lacking interpreter → real verdict; update 2 masking testsWP03[P]
T011Test: concurrent runs serialize; lock-wait ≠ run-timeoutWP03[P]
T012Document sub-agent long-gate contract in the implement-review skillWP03[P]
T013Thread project_root through manifest reader/writerWP04[P]
T014Serialize output_path relative + reconstruct absolute on readWP04[P]
T015Test: relative on-disk; legacy-absolute tolerance; keying invariantWP04[P]
T016Regenerate committed agent_profiles_manifest.json cleanWP04[P]
T017Issue-matrix blocker leads with schema-drift/column detailWP05[P]
T018Test: malformed column named, not all-issues-"Missing rows"WP05[P]
T019Bulk-edit: drop low-weight verbs from the triggered sumWP05[P]
T020Test: ordinary-refactor False; rename-all True; single-HIGH TrueWP05[P]
T021Pristine-vs-insufficient predicate near _substantive.pyWP06[P]
T022setup-plan returns scaffolded for pristine; blocked for insufficientWP06[P]
T023Handle new state in next engine + source prompt; regen agent copiesWP06[P]
T024Test: pristine→scaffolded; insufficient→blocked; substantive→successWP06[P]
T025Route move-task staging via authority path (reuse WP01 seam); no guard exemptionWP07
T026#2580: route _mt_persist_wp_file shell_pid via canonical writerWP07
T027Test: dual pin — STATUS_STATE byte-unchanged AND zero lane kitty-specs/WP07
T028Arch guard: IC-07 diff untouched status-bundling symbolsWP07
T029Clean primary route for legitimate solo-empty coord (surface_resolver)WP08[P]
T030Same-path regression: read surface == write placement == PRIMARYWP08[P]
T031Preserve warning's true signal (coord-with-lanes empty still warns)WP08[P]
T032Update surface_resolver coord-empty-warning testWP08[P]

Work Packages

WP01 — Allocator + move-task no-op-stable against runtime frontmatter (FR-001, #2580 helper)

Goal: The allocator's uncommitted-artifact check ignores the shell_pid/base_* WP frontmatter the claim itself wrote (sourced from the canonical WP_FIELD_ORDER), so N lanes batch-allocate with zero inter-allocation commits — while any body/substantive change still blocks. Provides the canonical helper WP07 reuses. Priority: P1 (MVP). Independent test: sequential N-lane allocation needs 0 commits.

  • ✅ T001 Canonical runtime-field accessor from frontmatter.py::WP_FIELD_ORDER (WP01)
  • ✅ T002 _drop_runtime_frontmatter_only_wp + wire into resolve_planning_artifact_staging (WP01)
  • ✅ T003 Test: runtime-only WP diff dropped; N-lane 0-commit allocation (WP01)
  • ✅ T004 Test: body/non-runtime change still blocks; no-op when auto_commit=True (WP01)

Prompt: tasks/WP01-allocator-runtime-frontmatter.md · ~230 lines · deps: none

WP02 — Analysis-report freshness normalizes pipe-table status cells (FR-002)

Goal: mark-status's pipe-table [D]/[P] cell churn no longer re-stales the analysis-report; a substantive tasks.md row change still stales it. Priority: P1. Independent test: mark-status→ implement on a pipe-table mission yields 0 spurious stale_analysis_report.

  • ✅ T005 Extend _normalize_tasks_md to canonicalize pipe-table status cells (WP02)
  • ✅ T006 Test: pipe-table [D]/[P] + mixed churn stays current (WP02)
  • ✅ T007 Test: substantive tasks.md row change still stales (true-positive) (WP02)

Prompt: tasks/WP02-freshness-pipe-table.md · ~170 lines · deps: none

WP03 — Pre-review gate runner: interpreter + contention + contract (FR-003, FR-004, FR-005)

Goal: The gate resolves the runner via uv run (so a uv-managed checkout returns a real verdict, not no_coverage→--force), is contention-safe under concurrent lanes (lock with its own timeout, decoupled from the subprocess timeout), and the sub-agent long-gate contract is documented. Priority: P1. Independent test: gate under a pytest-lacking interpreter returns a real verdict.

  • ✅ T008 Interpreter-resolution helper (uv run → sys.executable fallback) (WP03)
  • ✅ T009 Apply helper + contention lock (own timeout, decoupled from run timeout) (WP03)
  • ✅ T010 Test: pytest-lacking interpreter → real verdict; update 2 masking tests (WP03)
  • ✅ T011 Test: concurrent runs serialize; lock-wait ≠ run-timeout (WP03)
  • ✅ T012 Document sub-agent long-gate contract in the implement-review skill (WP03)

Prompt: tasks/WP03-gate-runner-hardening.md · ~300 lines · deps: none

WP04 — Manifest output_path repo-relative (FR-006)

Goal: The committed agent_profiles_manifest.json stores output_path repo-relative (in-memory + key stay absolute); the reader tolerates legacy absolute values with zero migration, so upgrade is cross-machine deterministic. Priority: P2. Independent test: upgrade on a second path → 0 diff.

  • ✅ T013 Thread project_root through manifest reader/writer (WP04)
  • ✅ T014 Serialize output_path relative + reconstruct absolute on read (WP04)
  • ✅ T015 Test: relative on-disk; legacy-absolute tolerance; keying invariant (WP04)
  • ✅ T016 Regenerate committed agent_profiles_manifest.json clean (WP04)

Prompt: tasks/WP04-manifest-relative-path.md · ~220 lines · deps: none

WP05 — Diagnostics papercuts: issue-matrix error + bulk-edit inference (FR-007, FR-008)

Goal: The issue-matrix blocker names the schema-drift column (not a misleading "Missing rows"); bulk- edit inference stops blocking on ordinary refactor verbs while genuine bulk edits — including single-HIGH- phrase — still trip. Priority: P2. Independent test: malformed-column error names the column; inference scoring cases pass.

  • ✅ T017 Issue-matrix blocker leads with schema-drift/column detail (WP05)
  • ✅ T018 Test: malformed column named, not all-issues-"Missing rows" (WP05)
  • ✅ T019 Bulk-edit: drop low-weight verbs from the triggered sum (WP05)
  • ✅ T020 Test: ordinary-refactor False; rename-all True; single-HIGH True (WP05)

Prompt: tasks/WP05-diagnostics-papercuts.md · ~230 lines · deps: none

WP06 — plan scaffold-block ergonomics (FR-009)

Goal: The first happy-path setup-plan scaffold write returns a distinct non-error state (scaffolded), not blocked; a populated-but-insufficient plan still returns blocked. The next engine + source prompt handle the new state; the specify side is already scaffold_only: success. Priority: P2. Independent test: fresh mission → first setup-plan returns non-blocked.

  • ✅ T021 Pristine-vs-insufficient predicate near _substantive.py (WP06)
  • ✅ T022 setup-plan returns scaffolded for pristine; blocked for insufficient (WP06)
  • ✅ T023 Handle new state in next engine + source prompt; regen agent copies (WP06)
  • ✅ T024 Test: pristine→scaffolded; insufficient→blocked; substantive→success (WP06)

Prompt: tasks/WP06-scaffold-block-ergonomics.md · ~250 lines · deps: none

WP07 — move-task coord-lane staging via authority path + #2580 (FR-010, #2580)

Goal: move-task on a coord-topology lane routes planning-artifact staging through the authority path so no manual git restore is ever needed, with NO commit_guard exemption and STATUS_STATE placement byte-unchanged; also routes the #2580 _mt_persist_wp_file shell_pid write through the canonical writer. Priority: P2. Coordination-aware (C-002). Independent test: coord move-task → dual pin holds.

  • □ T025 Route move-task staging via authority path (reuse WP01 seam); no guard exemption (WP07)
  • □ T026 #2580: route _mt_persist_wp_file shell_pid via canonical writer (WP07)
  • □ T027 Test: dual pin — STATUS_STATE byte-unchanged AND zero lane kitty-specs/ (WP07)
  • □ T028 Arch guard: IC-07 diff untouched status-bundling symbols (WP07)

Prompt: tasks/WP07-movetask-authority-staging.md · ~260 lines · deps: WP01

WP08 — Solo PR-bound coord mission routes empty-coord surface to primary (FR-011, #2533)

Goal: A solo PR-bound --start-branch mission whose coord worktree is legitimately empty resolves its status surface cleanly to PRIMARY (no split-brain warning / manual flatten), with the read surface proven to match the write placement. Consequence-only (the pr_bound⇒coord derivation stays; revisited in a follow-up). Priority: P2. Coordination-aware (C-002 WP07↔WP08). Independent test: solo-empty coord → primary, no warning.

  • □ T029 Clean primary route for legitimate solo-empty coord (surface_resolver) (WP08)
  • □ T030 Same-path regression: read surface == write placement == PRIMARY (WP08)
  • □ T031 Preserve warning's true signal (coord-with-lanes empty still warns) (WP08)
  • □ T032 Update surface_resolver coord-empty-warning test (WP08)

Prompt: tasks/WP08-solo-coord-surface-routing.md · ~200 lines · deps: none

Dependencies & Parallelization

  • WP01 is the MVP and unblocks WP07 (shared canonical runtime-field helper).
  • WP02–WP06, WP08 are fully independent and parallelizable (disjoint files).
  • WP07 depends on WP01; WP08 is disjoint from WP07 by file, coordinated by the C-002 boundary rule.

Suggested order: WP01 → (WP02, WP03, WP04, WP05, WP06, WP08 in parallel) → WP07.