Cy Create Tasks

Break down into tasks.

• Decompose implementation sections from the TechSpec into granular, independently implementable tasks.
• Each task MUST be independently implementable when all of its declared dependencies are met. No task may require undeclared work from another task. If two tasks share a tight coupling, either merge them or extract the shared piece into a dependency task.
• No circular dependencies. If task A depends on task B, task B must NOT depend on task A (directly or transitively).
• Each task must have: title, domain, type, scope, complexity, and dependencies.
• Assign complexity using these criteria:
  • low: Single file change, no new interfaces, no concurrency, straightforward logic.
  • medium: 2-4 files, may introduce a new interface or struct, limited integration points.
  • high: 5+ files, new subsystem or significant refactor, multiple integration points, concurrency involved.
  • critical: Cross-cutting change affecting many packages, high risk of regression, requires coordination with other tasks.
• When a task directly implements or is constrained by a specific ADR, include the ADR reference in the task's "Related ADRs" section under Implementation Details.
• Embed test requirements in every task. Never create separate tasks dedicated solely to testing.
• Follow the structure defined in references/task-template.md.
• Refer to references/task-context-schema.md for metadata field definitions.

Present task breakdown for interactive approval.

• Show all tasks with: titles, descriptions, complexity ratings, and dependency chains.
• Wait for user feedback before proceeding.
• If the user requests changes, revise the breakdown and present again.
• Iterate until the user explicitly approves.

Generate task files.

• Write _tasks.md as the master task list using this exact markdown table format:

  # [Feature Name] — Task List
  
  ## Tasks
  
  | # | Title | Status | Complexity | Dependencies |
  |---|-------|--------|------------|--------------|
  | 01 | [Task title] | pending | [low/medium/high/critical] | [task_NN, ... or —] |
  

• Write individual task files as task_01.md, task_02.md, through task_N.md.
• Task files use the task_ prefix without a leading underscore.
• Each file must start with YAML frontmatter containing status, domain, type, scope, complexity, and dependencies. Use dependencies: [] when there are no dependencies — do not omit the field.
• Task numbering must be sequential and consistent between _tasks.md and individual files.

Enrich each task file.

• For each task file, check whether it already has ## Overview, ## Deliverables, and ## Tests sections. If all three exist, skip enrichment for that file.
• Map the task to PRD requirements and TechSpec guidance.
• Spawn an Agent tool call to discover relevant files, dependent files, integration points, and project rules for this specific task.
• Fill ALL template sections from references/task-template.md. Every task file MUST contain each of the following sections — omitting any is a failure:
  • ## Overview: what the task accomplishes and why, in 2-3 sentences.
  • <critical> block: the standard critical reminders block (read PRD/TechSpec, reference TechSpec, focus on WHAT, minimize code, tests required).
  • <requirements> block: specific, numbered technical requirements using MUST/SHOULD language.
  • ## Subtasks: 3-7 checklist items describing WHAT, not HOW.
  • ## Implementation Details: file paths to create or modify, integration points. Reference TechSpec for patterns.
  • ### Relevant Files: discovered paths from codebase exploration with brief reasons.
  • ### Dependent Files: files that will be affected by this task with brief reasons.
  • ### Related ADRs: links to relevant ADRs if any exist, or omit subsection if no ADRs apply.
  • ## Deliverables: concrete outputs with mandatory test items and at least 80% coverage target.
  • ## Tests: specific test cases as checklists, split into unit tests and integration tests categories.
  • ## Success Criteria: measurable outcomes including "All tests passing" and "Test coverage >=80%".
• Reassess complexity based on exploration findings and update if changed.
• Update the task file in place with enriched content.
• If enrichment fails for one task, continue to the next and report all failures at the end.