SuperAPI AI Directory

to-issues

General & other

View repository

Updated 10-16 06:16

About this skill

Break work into tracer-bullet vertical slices — each cuts all layers end-to-end, demoable alone.

Run /aiops-setup if tracker config is missing.

Modes

  • Multi-session (PRD exists): full issue decomposition → issues/*.md published to tracker.
  • Single-session (no PRD, task_breakdown phase): lightweight → single tasks.md in .scratch/<slug>/, no tracker publish.

Process

  1. Context — conversation or fetched issue/PR from docs/agents/issue-tracker.md. For single-session, use tech-spec.md + NOTES.md as input.
  2. Explore (optional) — codebase, domain vocabulary, prefactor opportunities.
  3. Draft slices — title, blocked-by, user stories covered. Prefactor slices first.
  4. Quiz user — granularity, dependencies, merge/split. Iterate until approved.
  5. Publish — dependency order; ready-for-agent unless told otherwise. Don't close parent. For single-session, write tasks.md to .scratch/<slug>/.

Single-session two-phase breakdown (task_breakdown phase)

When in task_breakdown phase (single-session), run as two sub-phases with a human checkpoint between:

Phase 1 — parent tasks (3–5 items)

  • Draft coarse-grained parent tasks from tech-spec.md + NOTES.md
  • Each parent task is one demoable vertical slice
  • Present to user: confirm, reorder, merge, or split before proceeding

Phase 2 — sub-task expansion (only after user confirms parent tasks)

  • Expand each parent task into concrete sub-tasks with explicit blocked_by dependencies
  • Compute waves via topological sort
  • Write tasks.md with YAML frontmatter

Issue template

Parent — link if applicable.

What to build — end-to-end behavior, not layer checklist.

Acceptance criteria — checkboxes.

Blocked by — issue refs or "None".

tasks.md template (single-session)

Include YAML frontmatter for structured dependency tracking:

---
tasks:
  - id: t1
    title: "Add WorkspacePaths dataclass"
    blocked_by: []
  - id: t2
    title: "Wire workspace_paths() resolver"
    blocked_by: [t1]
  - id: t3
    title: "Update path helpers to use WorkspacePaths"
    blocked_by: [t1]
  - id: t4
    title: "Integration tests"
    blocked_by: [t2, t3]
waves:
  - [t1]
  - [t2, t3]
  - [t4]
---

# Tasks: <slug>

## Wave 1
- [ ] **t1**: Add WorkspacePaths dataclass — <end-to-end behavior>. Blocked by: None

## Wave 2
- [ ] **t2**: Wire workspace_paths() resolver — <end-to-end behavior>. Blocked by: t1
- [ ] **t3**: Update path helpers to use WorkspacePaths — <end-to-end behavior>. Blocked by: t1

## Wave 3
- [ ] **t4**: Integration tests — <end-to-end behavior>. Blocked by: t2, t3

Group tasks by wave in the body. Waves are computed from blocked_by via topological sort.