⚡ Do This First: Load Agent Profile

Before reading anything else, load your assigned agent profile with the /spk-charter-profile-load skill (profile: python-pedro, role: implementer). Then continue.

Work Package Prompt: WP01 – Research owned-checkout placement (#5882)

Objectives & Success Criteria

spec-kitty research accepts --owned-checkout <P> and, when run from inside an owned checkout, auto-adopts it (FR-001, FR-002).
For an owned run, research.md / data-model.md / the two research CSV stubs are scaffolded under owned.mission_dir; nothing is resolved, read, or written under repository_root/kitty-specs (FR-003, SC-001, SC-002).
An invalid owned target refuses via emit_owned_refusal before any mkdir/copy (NFR-003).
The non-owned path is byte-for-byte behaviour-identical; tests/specify_cli/cli/commands/test_research_read_surface.py stays green (NFR-001).
Ownership resolved only through the existing _owned_checkout / core.owned_mission seam; no new resolver (NFR-002, C-001). ruff/mypy clean on changed files.

Context & Constraints

Grounding + provenance: ../analysis-report.md (§2 Surface A) and ../plan.md IC-01.
Reproduction (already RED on the base): tests/integration/test_owned_research_placement.py — test_research_accepts_owned_checkout_and_scaffolds_into_owned and test_research_from_owned_cwd_does_not_redirect_into_dirty_primary, built on the canonical owned_checkouts / stale_root_copy / r_snapshot fixtures.
Root cause (origin/main research.py): research(...) (170-177) exposes only --mission/--force; repo_root = find_repo_root() (183) + get_project_root_or_exit (188) follow the owned worktree's .git pointer to the PRIMARY; the three _read_mission_dir_or_exit(repo_root, slug, kind) calls (212 STATUS_STATE, 248 RESEARCH, 261 FINALIZED_EXECUTION_PLAN) key the placement seam on the primary root with no owned=.
Pattern to MIRROR (do NOT invent a new resolver, C-001): mission_record_analysis.py (option → resolve_owned_or_refuse/resolve_owned_or_adopt → repo_root = owned.owned_root → placement_seam(repo_root, slug, owned=owned)) and mission_setup_plan.py (owned → read dirs resolve to owned.mission_dir). Use resolve_owned_or_adopt so a flagless run from inside an owned checkout adopts it and a run from the primary keeps repo-root behaviour.
Charter: C-001 single resolver / no core/paths.py change; DIRECTIVE_024 locality; smallest-viable-diff.

Subtasks & Detailed Guidance

Subtask T001 – ATDD: owned research placement reproduction (RED first)

Purpose: Pin the containment invariant before the fix.
Status: the reproduction file already exists and is RED on the base (No such option: --owned-checkout; flagless run resolves the primary). Confirm RED on the WP base and record the failure output.
Validation: .venv/bin/python -m pytest tests/integration/test_owned_research_placement.py -q → both tests RED.

Subtask T002 – Add `--owned-checkout` to `research`; resolve ownership once

Purpose: Resolve ownership at the CLI boundary and carry the fact down.
Steps: Add owned_checkout: OwnedCheckoutOption = None to research(...). After repo_root = find_repo_root(), resolve once via resolve_owned_or_refuse(repo_root, owned_checkout, mission, cwd=Path.cwd(), allowed_topologies=LIFECYCLE_OWNED_TOPOLOGIES, json_output=False, envelope=...) (research has no --json; render the refusal on stderr). When owned is not None, set project_root = repo_root = owned.owned_root.
Files: src/specify_cli/cli/commands/research.py.

Subtask T003 – Thread `owned=` through the placement seam + validate owned plan

Purpose: Make every mission-dir resolution and the plan-filled gate honour the owned fact.
Steps: Give _read_mission_dir_or_exit an optional owned: OwnedCheckout | None = None and forward it as placement_seam(repo_root, slug, owned=owned).read_dir(kind); pass owned=owned at all three call sites (STATUS_STATE, RESEARCH, FINALIZED_EXECUTION_PLAN). The RESEARCH planning_dir is where the four assets are written; the FINALIZED_EXECUTION_PLAN dir is where validate_plan_filled reads plan.md — both now land on owned.mission_dir. Keep _write_research_asset unchanged (it already writes under planning_dir).
Files: src/specify_cli/cli/commands/research.py.

Subtask T004 – Validate green (owned + non-owned read-surface)

Steps: .venv/bin/python -m pytest tests/integration/test_owned_research_placement.py tests/specify_cli/cli/commands/test_research_read_surface.py tests/research -q (owned GREEN, non-owned read-surface unchanged). make format-check-files FILES=src/specify_cli/cli/commands/research.py; ruff check src/specify_cli/cli/commands/research.py; mypy on the changed file (attribute only NEW diagnostics vs the #5917 baseline).

Definition of Done

Repro RED on base → GREEN on final commit; non-owned research behaviour unchanged; refusals fire before writes; lint/type clean (no NEW mypy vs #5917).

Reviewer Guidance

Confirm no find_repo_root/get_project_root_or_exit result reaches the placement seam for an owned run (all three _read_mission_dir_or_exit calls thread owned=); confirm resolve_owned_or_refuse/_adopt (not a new resolver); confirm R-immutability holds against a stale primary copy.

Spec Kitty

Kanban Board