Spec Kitty

└─ kitty-specs
   └─ Execution-Context Unification

Mission Run:

📚 Docs ↗

Quickstart / Validation — Execution-Context Unification (01KTPKST)

Validation scenarios that prove the mission's success criteria. The parity ratchet (IC-08) is the primary regression guard; the rest are targeted repros of the drained issues.

SC-1 — Parity ratchet (the regression guard)

EXTEND tests/architectural/test_execution_context_parity.py (do not fork). Assert that resolving the context from the primary checkout and from a lane/coord CWD yields identical answers for every fragment, across specify → plan → tasks → analyze → implement → review → status.

pytest tests/architectural/test_execution_context_parity.py -q
# Expect: identical fragment values from both CWDs; zero split; clean trees.

Add a flattened-topology synthetic fixture (no separate coord branch) so C-001 is proven: the flattened mission must show CommitTarget.kind == flattened, coordination_branch is None, status_read_dir == status_write_dir.

SC-2 — Paused-mission blockers do not reproduce

The two failures that paused 01KTNWFC must be structurally precluded after IC-05:

# record-analysis must not deadlock on coord residue (#1814)
spec-kitty agent mission record-analysis --mission <fresh-mission> --input-file <report.md> --json
# implement-claim must not block on planning-artifact branch-split (#1816)
spec-kitty agent action implement WP01 --mission <fresh-mission> --agent claude:opus:implementer:implementer

Both resolve placement via the ArtifactPlacementFragment — no primary-dirty-tree vs coord-owned conflict.

SC-3 — Each folded issue's repro passes (or is structurally precluded)

Issue	Repro check
#1737	`status_transition` reads the carried StatusSurfaceFragment; grep confirms no independent coord-path derivation remains
#1357	concurrent `CoordinationWorkspace.resolve` is lock-serialized (no race materializing divergent surfaces)
#1572	status visible identically from primary and coord CWD
#1764	analysis-report not falsely stale across CWDs
#1789/#1062	see SC-5
#1735/#1771	retrospect reads/writes the canonical surface (no primary-only read, no gitignored write)
#1736/#1770	merge consumes the context (PATH/env, baking, JSONL)
#1816/#1814	see SC-2

SC-4 — One resolution path (C-005 enforcement)

# No second context resolver, no forked parity test, no resurrected duplicate.
rg -n "candidate_feature_dir_for_mission" src/        # expect: gone (folded into _read_path_resolver)
rg -n "def .*parse.*worktree" src/specify_cli/workspace/root_resolver.py  # expect: gone (deleted)
rg -n "EventLogWriteTarget|StatusReadSource|read_wp_lane_actor" src/      # expect: gone (IC-09)
ruff check . && mypy <changed paths>                  # zero issues (NFR-002)

SC-5 — Long rebase does not clobber status (#1789)

# On a mission branch with a daemon/dashboard active, run a long rebase.
# materialize_if_stale must NOT re-materialize tracked status mid git-op.
git rebase <base>   # completes with no status-file clobber; tree clean afterward

NFR checks

pytest tests/architectural/test_no_legacy_terminology.py   # Terminology Canon (pre-push, CI-only gate)
pytest tests/architectural/test_shared_package_boundary.py
# NFR-005: changed-path LOC trends DOWN (report delete vs add in the WP handoff)

Cleanup discipline

branches (known E2E leak). Clean both surfaces after the test run.

The flattened synthetic fixture must not leak test-feature- missions or kitty/mission-test-feature-