Pm Autoscope

• Set story dependencies: If a story depends on another story being done first, use depends_on

Incremental Mode (existing stories need tasks)

Triggered when stories exist but some have no tasks. This is the primary use case.

Workflow:

1. pm_auto_scope returns the list of undecomposed stories with their bodies
2. Loop through each story: a. Call pm_scope(story_id) to get full decomposition context b. Propose 2-6 tasks for the story based on its content c. Present the task list to the user for quick approval (yes/edit/skip) d. Create approved tasks with pm_create_task
3. After all stories are scoped, show summary:
   • N stories scoped
   • N tasks created
   • Total points assigned
4. Suggest next steps: /pm board to see available work or /pm-plan for sprint planning

Guidelines

• Keep task titles as verb phrases: "Add authentication middleware", "Write unit tests for parser"
• Each task should be 1-5 points (completable in one session)
• Tasks should be independently testable
• First task in a story sets up the foundation
• Last task handles integration/cleanup
• Don't over-decompose — 2-6 tasks per story is the sweet spot

Dependencies: depends_on

Intra-Story Dependencies (within a story)

• Create implementation tasks BEFORE test tasks — implementation tasks get lower IDs
• Test tasks MUST set depends_on pointing to the implementation task(s) they verify
• This ensures the board shows implementation tasks first, and pm_grab picks them in the right order
• Example: if task -1 is "Add schema" and task -2 is "Test: schema", then -2 must have depends_on: ["US-PRJ-X-1"]

Cross-Story Dependencies (between stories)

• Tasks can depend on tasks from other stories: Use when work in one story requires something from another
• Stories can depend on other stories: Use when a story can't start until another is complete
• Example: US-PRJ-2-1 (frontend integration) depends on US-PRJ-1-3 (API endpoint) from a different story

Cross-Story Patterns

When scoping related stories, identify cross-story dependencies:

• API-first: Backend story tasks should be done before frontend integration tasks
• Shared infrastructure: Common utilities story should complete before dependent feature stories
• Data flow: Data model tasks should precede tasks that consume that data

Batch Creation with Dependencies

Use pm_create_tasks (batch) to create all tasks at once:

• List implementation tasks first in the array
• List test tasks last with depends_on set
• For cross-story deps, reference existing task IDs from other stories

Forcing a mode

Users can force a specific mode:

• /pm autoscope full — force full codebase scan even if stories exist
• /pm autoscope incremental — force incremental even if no stories exist