Task action items¶

Per-action-item rows attached to a parent task. Each item carries its own risk profile + required contexts + definition_of_done. Safety rule: items with authorship='agent_unapproved' cannot be executed by the agent -- is_executable enforces this.

Details¶

Per-action-item rows attached to a parent task. Module: work_buddy/obsidian/tasks/action_items.py. Schema: task_action_items table + task_metadata.current_action_item_id column.

Schema¶

Per-item: id, task_id (FK), sequence (UNIQUE per task), description, state (pending | in_progress | done | skipped), risk_profile_json, agent_required_contexts, user_required_contexts, definition_of_done, authorship, completed_at, handoff_package_path, deleted_at, created_at, updated_at.

The authorship column is a three-valued enum (work_buddy.threads.enums.Authorship):

Value	Meaning
`USER`	User wrote the item directly. Executable.
`AGENT_APPROVED`	Agent proposed it; user clicked Approve. Executable.
`AGENT_UNAPPROVED`	Agent proposed it; user has not yet approved. NOT executable.

CRUD entry points¶

create(task_id, *, description, sequence=None, ...) -- auto-assigns next sequence when None.
get(item_id), list_for_task(task_id, *, include_done=True), update(item_id, **fields) (sentinel discipline like store.update), delete(item_id) (soft -- sets deleted_at; see restore to undo).
approve(item_id) -- flips authorship from agent_unapproved to agent_approved (the canonical way for the user to admit an agent-proposed item to execution).
set_current(task_id, item_id) -- points task_metadata.current_action_item_id (None to clear).

Safety rule¶

is_executable(item) returns True iff:

authorship is USER or AGENT_APPROVED, AND
state is not in {done, skipped} (terminal exclusion), AND
deleted_at is NULL.

Every executor / Resolution Surface accept-action button / develop-at-pickup write step MUST call is_executable before running an item.

position_in_task¶

position_in_task(item) -> (1-based-index, total) powers the master-list "step N of M" badge.