subagents-tasks

Summary

Convert PRD/issue/todo/conversation into an Agents task breakdown markdown files with explicit sub-agent checkpoint workflow. The output file should be structured for iterative execution, with clear next steps and signals for sub-agents to pick up work in sequence.

When to use

• You have a PRD, issue list, todo list, or raw discussion and need a structured, execution-ready plan.
• You want to make work decomposable into cycles.
• You need clear outputs for engineering, design, or PM handoff.

Input options

• PRD text (goals, users, metrics, scope).
• GitHub issue list, Jira stories, or plain checklist.
• Todo list with tasks.
• Conversation notes (meeting, brainstorming, chat log).

Output format

• Output file must be created as a filesystem artifact.
  • Default output directory: /tmp/ unless the workspace contains a .tasks folder or the user explicitly requests a different output location. If a .tasks folder exists, write outputs into .tasks/. If the user specifies a path, use that instead.
  • File path pattern: ${descriptive-name}-${timestamp}.md (agent should derive a concise descriptive name).
  • Use the agent filesystem action (e.g., create_file) to write the final markdown into the chosen directory using the pattern directory/${descriptive-name}-${timestamp}.md.
  • Ensure the target directory exists (create it if necessary) before writing.
  • Include the generated file path explicitly in the final summary.
• Separation: produce one plan folder named ${descriptive-name} containing a tasks.md control file and individual task{n}.md files (required).
  • Output folder: create ${output_dir}/${descriptive-name}/.
    • Default output directory: /tmp/ unless the workspace contains a .tasks folder or the user explicitly requests a different output location. If a .tasks folder exists, prefer and use .tasks/ inside the workspace.
  • Control file: ${descriptive-name}/tasks.md — contains the master task list, overall idea, and each task's status (Pending, In progress, Done).
  • Task files: ${descriptive-name}/task1.md, ${descriptive-name}/task2.md, ... — each file contains full task context, acceptance criteria, dependencies, estimated effort, and an explicit Next sub-agent command.
  • The control tasks.md must reference each task file with workspace-relative paths and include Sub-agent workflow sections linking to the task files.
  • Use the agent filesystem action (create_file) to create the folder, tasks.md, and each taskN.md. Ensure directories exist before writing and include all created file paths in the final summary.
• The file must be sub-agent checkpoint friendly: include an ordered work queue and explicit Next sub-agent status markers.
• Example output skeleton:
  • ## Context
  • ## Goal + metrics
  • ## Hypothesis
  • ## Experiments / increments
  • ## Acceptance criteria
  • ## Blockers/risks
  • ## Next steps
  • ## Sub-agent workflow (Pending, In progress, Done, Signals)

Step-by-step process

1. Identify source type
   • Ask: "Is this a PRD, issue list, todo list, or conversation?"
   • If ambiguous: request the most structured form available.
   • If user submits unstructured raw text: ask for the most concise, structured reframing (e.g., PRD/issue/todo format).
2. Extract top-level problem and success criteria
   • User problem, business objective, and measurable KPIs.
   • Constraints (timeline, budget, tech limits, compliance).
3. Normalize scope
   • Must-have vs nice-to-have.
   • Key user scenarios and personas.
4. Derive candidate blocks
   • Problem statement
   • Hypothesis (if unknown, propose)
   • Experiment(s) / increment(s)
   • Data/metrics for validation
   • Dependencies, blockers, risk mitigations
5. Split into action items
   • Concrete tasks (design/dev/test/launch)
   • Estimated effort (relative: small/medium/large)
   • If estimate is large, consider breaking down further or ask the user for more details to refine.
   • Each task is handled by a sub-agent (task-specific processing)
6. Build the Sub Agent Iterative Ready file
   • ## Context
   • ## Goal + metrics
   • ## Hypothesis
   • ## Experiments / increments
   • ## Acceptance criteria
   • ## Blockers/risks
   • ## Next steps
   • Build artifacts inside ${descriptive-name}/:
     1. tasks.md (control): ordered master work queue, global metadata, and the Sub-agent workflow with Pending, In progress, Done, and explicit links to each taskN.md.
     2. taskN.md files: each contains full context, dependencies, acceptance criteria, estimated effort, and an explicit Next sub-agent command instructing the downstream sub-agent how to execute and how to update both the task file and tasks.md.
   • ## Sub-agent workflow (in tasks.md)
     • ### Pending (list of taskN.md links)
     • ### In progress (single current task entry with assignee/sub-agent and link to the taskN.md)
     • ### Done (completed task entries with links to taskN.md and short status notes)
     • ### Signals (machine-friendly flags or short commands for sub-agents; e.g., run-task: ${descriptive-name}/task2.md)
   • Each taskN.md must include an explicit candidate command text suitable for downstream sub-agents, e.g., "Execute step 2, run tests, then update task2.md and set tasks.md status to Done."
7. Validation checklist
   • Problem + outcomes are clear
   • Success criteria measurable
   • Scope is clearly surfaced
   • Tasks actionable and upper bound ready
   • No missing dependencies
   • Sub-agent state transitions are defined in the file
   • Control file created and references all task files
   • Each task has a standalone markdown file with context and Next sub-agent command
   • All file paths included in final summary
   • tasks.md created inside ${descriptive-name}/ and references all taskN.md files
   • Each taskN.md is created inside ${descriptive-name}/ with context and Next sub-agent command
   • All created file paths included in final summary

Decision points / branching logic

• If input is PRD/issue/todo, prefer structured extraction and minimal clarifying questions.
• If input is conversation, ask clarifying questions to confirm:
  • objective, KPI, target audience, deliverables.
• If success criteria are missing, generate candidate metrics and ask for confirmation.

Quality criteria

• Execution-ready with clear handoff artifacts.
• Includes at least one hypothesis per feature area.
• Contains explicit acceptance criteria and test signals.
• Presents one-liner update for standups.

Prompt examples

• "Convert this PRD into an iterative ready plan with hypothesis, experiments, and a next-step checklist."
• "I have this issue list: [...]; convert to a single tasks file."
• "From this meeting transcript, draft an iterative ready plan and identify missing info."