2lab-ai/plan-new-task

Plan New Task

Goal

When there is no meaningful unfinished work in docs/*/trace.md, proactively propose new feature work based on completed features and project context, then create structured STV artifacts using stv:new-task.

Sizing Rubric (expected code change, added + deleted)

| Tier | Lines | Example | |--------|--------|--------------------------------------------| | tiny | ~5 | Config values, constants, string literals | | small | ~20 | One function, one file, local refactor | | medium | ~50 | Multiple files, interface changes | | large | ~100 | Cross-cutting concerns, schema migrations | | xlarge | ~500 | Architecture shift, framework replacement |

Workflow

1. Confirm backlog status

   • Glob for docs/*/trace.md in the project
   • Read Implementation Status from each trace.md
   • Confirm no meaningful unfinished scenarios exist
   • If tiny/medium leftovers exist, list them separately as carryover

2. Review completed work

   • Scan docs/*/ directories for completed features
   • Read spec.md and trace.md summaries
   • Identify patterns: what was built, what's missing, what's the natural next step
   • Check git log for recent work context

3. Propose next work

   • Suggest two to four candidate features based on:
     • Completed feature history (natural follow-ups)
     • Visible gaps in the codebase
     • User context and project goals
   • Tie each idea to recent work or visible gaps
   • Estimate size tier for each proposal

4. Present options

All trace scenarios are complete. Based on project history, here are good next features:

1) {Idea A} ({estimated tier})
   - {one line reason, tied to recent work}

2) {Idea B} ({estimated tier})
   - {one line reason}

3) {Idea C} ({estimated tier})
   - {one line reason}

Carryover (tiny leftovers, optional):
- docs/{feature}/trace.md — Scenario {n}: {title}

Pick a number or describe a different feature.
I will then break it down with stv:new-task.

5. Plan with stv:new-task

   • After user picks an idea, invoke Skill(skill="stv:new-task")
   • stv:new-task handles: intent understanding → stv:spec → stv:trace → scenario task list

6. Hand off

   • After spec + trace are created, suggest using stv:do-work to execute

Integration

• Use stv:what-to-work to decide when to enter this flow (it routes here when backlog is empty/small)
• Use stv:new-task for feature decomposition into spec + trace
• Use stv:do-work after trace scenarios exist