Split Story Checklist

STOP - Read before ANY tool call

1. Read {{GUARDRAILS_PATH}}
2. Read _shared/sdd-artifacts/SESSION.md; load session-state for $Cwd
3. If the relevant gate is not approved: STOP - ask user (pt-BR) - do NOT Write/Shell
4. SDD/develop skills: after ONE step/task, STOP session - handoff only
5. This skill body is English; user-facing prompts may be (pt-BR)

Step -1 - Gate check (report in chat before continuing)

Gate check:
[ ] guardrails.mdc read
[ ] SESSION.md read; session-state loaded
[ ] PIPELINE.md read (SDD skills only)
[ ] User confirmed current action (sim)
-> If any unchecked: STOP

Skill: split-story-checklist

Trigger

Invoke when the user asks for: /split-story-checklist, break down tasks, /split-story-checklist, or after refine-story.

Input (one of):

Prerequisite: content includes structured Steps (or Bug Suggested fix). If missing, hand off to /refine-story.

Outcome

In the target workspace, a grouped checklist (backend / frontend / tests) with dependency-aware order (dependency-aware task checklist style).

Output altitude (RN01 / REQ-005): emit SMART engineering tasks only — never one US/TS per file, class, or script (anti-task-shatter.md). Keep task-shaped titles as checklist rows under the parent story; do not invent new USnn/TSnn folders from file names.

Persistence (prefer in order):

1. features/NNN-slug/USnn/REFINE/tasks.md - default when story folder exists
2. features/NNN-slug/USnn/TASKS.md - only if user explicitly asks for flat beside STORY.md (do not invent both)
3. Shortcut: docs/implementation-tasks/<slug>.md (legacy alias docs/sdd-developation-tasks/ still accepted)

If both REFINE/tasks.md and TASKS.md already exist: update REFINE/tasks.md and note the duplicate in chat (do not fork content into both).

TASKS complexity gate (REQ-004): Read FEATURE Complexity. If trivial (small) → do not Write TASKS / REFINE/tasks.md only to satisfy a gate — report that TASKS is not required. If medium or complex → TASKS checklist is required before handoff. Details: CHANGE-CONTRACT.md + references/complexity-policy.md.

No creation of external work items; no mandatory corporate workflow checklists.

Lazy-load

Never by default: do not preload all references/*.md or all backlog-item-types. Load one section per Process step (SKILL-REFERENCE-RETRIEVAL.md). Load anti-task-shatter.md before Write when titles look verb+file/class/script.

Reference routing

Process

Step -1b - Caveman Mode (Full cap)

1. Read {{SDD_ROOT}}/preferences.json (create { "caveman_mode": false, "caveman_level": "full" } if missing).
2. If caveman_mode is false: continue without compression.
3. If true: load {{TOOLKIT_ROOT}}/skills/_shared/caveman/CAVEMAN.md; apply Full participation cap + prefs caveman_level (Lite skills never escalate); show once: [Caveman] Modo ativo (respostas compactas, level={effective}). Digite caveman off para desativar.
4. Honor caveman on|off|status|lite|full|ultra (and stop caveman / normal mode) during the session.
5. Auto-Clarity + never-compress gates/drafts/paths per CAVEMAN.md.

0. Workspace and source

1. Confirm target repository.
2. Load refined content from feature story, path, chat, or paste.
3. Extract steps from Steps / Suggested fix (references/parsing.md).
4. Read sibling FEATURE.md Complexity when under features/NNN-slug/ (CHANGE-CONTRACT.md TASKS policy).

If no steps found, stop and suggest /refine-story.

If Complexity is trivial (small): STOP Write of TASKS — tell the operator (pt-BR) that TASKS is not required for trivial/small; offer optional checklist only if they insist.

If Complexity is medium or complex (or unknown and story looks multi-step): proceed to build the required TASKS file.

1. Documentation language (blocker before Write)

Ask once:

Language for the tasks file - pt-BR or English?

Record in the output file header. Paths stay English.

2. Group steps (deps + topology)

Apply references/topology-grouping.md:

• Honor explicit Depends on: edges first (Kahn / levels)
• Then layer / repo heuristics
• Mark steps safe to run in parallel when no edge between them
• Max 5 implementation groups - merge adjacent if needed
• Test steps go to dedicated test groups

3. Build checklist file

Write preferred path under the story folder (or shortcut) using references/output-template.md - honor Persistence order above (REFINE/tasks.md default):

• Implementation groups with - [ ] per original step (preserve titles and dependencies)
• Separate Tests section when test steps exist
• Execution order with critical path + parallel waves
• Optional Before PR neutral checklist (user may omit)

Each row is a SMART task under the existing story — not a new US/TS. If a step title is verb+file/class/script, keep it as a task row; do not promote it to a story folder (anti-task-shatter.md / RN01).

Do not inject fixed corporate tasks (AI tags, manual sign-off checklist, Sonar boilerplate as mandatory rows).

4. Summarize in chat

Show group names, dependency waves, output path, and suggested next skills.

5. Handoff

Must not

Also enforce references/exclusions.md. Handoff boundary: references/sdd-handoff.md. PLAN path: references/plan-resolution.md.

• Create or update external tracker cards via external work-item APIs
• Add fixed "workflow" tasks (manual sign-off checklist, Datadog, SDD/DevAI tags) unless the user explicitly requests a custom section
• Assume toolkit repo paths during consumer runs
• Write the file before the language question
• Emit one US/TS per file, class, or script (RN01) — output stays SMART tasks under the parent story (anti-task-shatter.md)
• Create new USnn/TSnn folders from checklist breakdown

Handoff examples

/sdd-spec

/orchestrate-analyze

/sdd-develop - features/004-export/US01/PLAN/PLAN_004_export.md - Step 1