kiss-wbs-decompose

name	kiss-wbs-decompose
description	Decomposes the project plan's Work Breakdown Structure (WBS) into a set of feature directories, writes a stub spec.md for each WBS leaf node, and updates .kiss/feature.json to point at the first feature. Acts as the bridge between programme-level planning (kiss-project- planning) and feature-level specification (kiss-specify). Use when a Waterfall project plan exists and the team needs to fan out the WBS into individual feature folders so that kiss-specify, kiss-plan, and kiss-implement can be run feature by feature.
compatibility	Requires kiss project structure with .kiss/ directory
metadata	{"author":"github-kiss","source":"agent-skills/kiss-wbs-decompose/kiss-wbs-decompose.md"}
user-invocable	true
disable-model-invocation	false

User Input

$ARGUMENTS

Consider the user input before proceeding (if not empty). If the user specifies a WBS section or level (e.g. "only Phase 2"), scope the decomposition to that subset.

Audience and tone (interactive mode)

When KISS_AGENT_MODE=interactive (the default), assume the user has limited technical background and limited domain knowledge. Run this skill as a guided questionnaire:

• One question at a time. No walls of questions.
• Yes / no first. Phrase so yes, no, not sure, or skip is a valid answer.
• Translate jargon, don't strip it. Explain WBS, leaf node, feature scope on first use.
• Choices, not blank fields. Offer lettered options (A/B/C/D). Always include "Not sure — sensible default".
• Always recommend. State which option you would pick and why.
• Show, don't ask. When the WBS is already in the plan, propose the parsed feature list and ask for confirmation.
• not sure / skip triggers a sensible default, marked "(default applied)" and a WBSDEBT entry.

When KISS_AGENT_MODE=auto, skip the questionnaire and log decisions.

Inputs

• .kiss/context.yml
• {context.paths.docs}/project/project-plan.md — WBS section (required; if absent, prompt user to run /kiss.project-planning first, then return here)
• {context.paths.docs}/project/project-plan.extract — structured KEY=VALUE from the planning skill
• {context.paths.specs}/ — scanned for existing feature directories to avoid duplicates

Outputs

• One new feature directory per WBS leaf: {context.paths.specs}/<NNN>-<slug>/spec.md — stub spec with the WBS item's title, work package ID, and parent path pre-filled. The stub is intentionally minimal; /kiss.specify refines it.
• {context.paths.docs}/project/wbs-index.md — master table mapping WBS ID → feature directory → status
• {context.paths.docs}/project/wbs-index.extract — companion KEY=VALUE ledger (WBS_LEAF_COUNT, FEATURE_DIRS_CREATED)
• {context.paths.docs}/project/wbs-debts.md — WBS items that could not be auto-decomposed (WBSDEBT-NN: …)

Context Update

Does not mutate .kiss/context.yml directly. Writes .kiss/feature.json to point at the first generated feature directory, so the next /kiss.specify picks it up automatically.

Handoffs

• /kiss.specify refines each stub spec into a full specification.
• kiss-srs aggregates all specs produced from the stubs into the SRS.
• kiss-traceability-matrix uses the WBS IDs from wbs-index.md as a cross-reference column.

AI authoring scope

Does:

• Parse the WBS section of project-plan.md to find all leaf nodes (items with no sub-items, or items at the user-confirmed decomposition level).
• Generate a slug (2-4 words, kebab-case) and a sequential prefix (NNN) for each leaf.
• Write a stub spec.md from templates/wbs-spec-stub.md with the WBS ID, title, parent WBS path, and a reminder to run /kiss.specify.
• Skip directories that already exist (idempotent).

Does not:

• Fill in full functional requirements — that is /kiss.specify's job.
• Modify project-plan.md.
• Delete or overwrite existing spec directories.

Outline

1. Load plan — read project-plan.md. If not found, stop and prompt the user to run /kiss.project-planning first.

2. Parse WBS — extract all items from the WBS section. In interactive mode, show the parsed leaf list and ask: a. Are these the correct leaf nodes to decompose? (yes / edit) b. Should summary/phase nodes also get stubs? (default: no) c. Should the sequential prefix restart from the current max existing number in specs/? (default: yes)

3. Check for duplicates — scan {context.paths.specs}/ for directories whose names contain the same slug. Skip any matches and note them.

4. Generate stubs — for each confirmed leaf node:

   • Compute next available NNN prefix.
   • Generate slug from WBS item title.
   • Write {paths.specs}/<NNN>-<slug>/spec.md from the stub template.

5. Write wbs-index.md — one row per leaf: WBS ID | Title | Feature Dir | Status (Stub / In Progress / Done).

6. Write .kiss/feature.json pointing at the first stub.

7. Summary — print:

   • N stubs created, M skipped (already exist).
   • wbs-index.md location.
   • Next action: run /kiss.specify for each feature in order.

Usage

<SKILL_DIR> = the integration's skills root.

WBS_DECOMPOSE_LEVEL=3 \
WBS_INCLUDE_SUMMARIES=false \
  bash <SKILL_DIR>/kiss-wbs-decompose/scripts/bash/decompose-wbs.sh --auto

Answer keys

Key	Meaning	Default
WBS_DECOMPOSE_LEVEL	Max WBS depth to treat as leaf	3
WBS_INCLUDE_SUMMARIES	true / false — also stub summary nodes	false
WBS_PREFIX_RESET	true / false — restart NNN from max existing	true