axon-ivy

name	axon-ivy
description	Master router skill for all requirement clarification, create story, development, review, and testing tasks. Detects user intent and orchestrates the correct path using specialized skills and parallel subagents.

When to Use

Use this skill as the single entry point whenever the user wants to do anything in an Axon Ivy project. It routes to the right path automatically.

---

Step 0: Check for Existing Ivy Project

Before doing anything else, search for .ivyproject files to locate existing Axon Ivy projects within this codebase only:

1. Use Glob with pattern **/.ivyproject scoped to the current working directory, skipping .claude/, .github/, and .vscode/ folders
2. For each .ivyproject found, read it to extract the name field, then locate and read the sibling pom.xml for groupId and artifactId
3. Present findings to the user in a short summary, e.g.:

   Found project: invoice-parser
   Group ID:      com.example
   Artifact ID:   invoice-parser
   Location:      invoice-parser/
   

If NO .ivyproject is found: automatically run Path A — Project Bootstrap first (no prompt needed), then — once Path A is fully complete — continue to the path that matches the user's original intent (Step 1 below). Do NOT skip ahead to requirements or stories before the project is scaffolded.

If one or more projects ARE found: present the summary, then proceed to Step 1.

---

Step 1: Detect Intent

Read the user's request and match it to one of the 4 paths:

If user says...	Path
"new project", "create project", "init", "scaffold"	A — Project Bootstrap
"implement story N", "work on story", "continue story N"	B — Story Pipeline
"build feature", "new feature", "create workflow", "add functionality"	C — Feature Builder
"verify", "review", "check my changes", "run checks", "quality gate"	D — Verify All

If intent is ambiguous, use your best judgment — freely decide which path fits the user's goal most naturally. State your choice in one line before starting, e.g.: → Routing to Path B: Story Pipeline

---

Path A: Project Bootstrap

Goal: Scaffold a new project that is ready for development.

Sequential steps in main context:

1. Run skill axon-ivy-init — scaffold project from template, run Maven build
2. After scaffold completes, spawn 2 parallel subagents:
   • config-setup — follow axon-ivy-variable-config + axon-ivy-user-role-config: create config/variables.yaml, config/roles.yaml, config/users.yaml
   • data-scaffold — follow axon-ivy-data: create base data classes in dataclasses/
3. Wait for both subagents to finish
4. Run skill axon-ivy-custom-fields if the project needs task/case metadata fields

Done when: project builds cleanly and base config files exist. If Path A was triggered as a prerequisite (i.e., no .ivyproject existed when the user asked for a feature or story), immediately continue to the originally intended path — do not stop here.

---

Path B: Story Pipeline

Goal: Implement a story end-to-end and verify it passes all checks.

Steps:

1. Read the story file [project]/documents/stories/story_NN.md

2. Run skill axon-ivy-implement-story — implement all parts sequentially

3. After all parts are done, spawn 3 parallel verification subagents:

   Subagent	Follows	Scope
   process-verify	axon-ivy-process-verify	All .p.json files modified in this story
   cms-verify	axon-ivy-cms-verify	All cms_*.yaml files modified in this story
   ui-verify	axon-ivy-primefaces-verify	All .xhtml files modified in this story

4. Collect results from all 3 subagents

5. If any subagent reports failures → fix the issues, then re-spawn only the failing subagent(s)

6. Once all pass, run axon-ivy-verify-story in main context for final acceptance criteria check

7. Mark story checkboxes [x] on pass

---

Path C: Feature Builder

Goal: Turn a plain feature description into a fully implemented and verified feature.

Steps:

1. Run skill axon-ivy-requirements-creation — produce structured requirements doc + ordered story files
2. Present the story list to the user; confirm the implementation order
3. Spawn a story-planner subagent to identify which stories can run in parallel (e.g., CMS stories alongside process stories)
4. For each story (or parallel batch), execute Path B: Story Pipeline
5. After all stories pass, run Path D: Verify All as a final gate

---

Path D: Verify All

Goal: Run the full verification and review suite on recent changes and produce a consolidated report.

Spawn 6 parallel subagents simultaneously:

Subagent	Follows	Scope
process-verify	axon-ivy-process-verify	All modified .p.json files
cms-verify	axon-ivy-cms-verify	All modified cms_*.yaml files
ui-verify	axon-ivy-primefaces-verify	All modified .xhtml files
story-verify	axon-ivy-verify-story	Current story acceptance criteria
java-review	review-java	All modified .java files
js-review	review-javascript	All modified .js / .ts files

Skip subagents where no matching files exist in the current changeset.

Wait for all, then output a consolidated report:

✓ Process verify:     PASS  (N files checked)
✗ CMS verify:         FAIL  — cms_en.yaml line 42: key mismatch
✓ PrimeFaces verify:  PASS  (N files checked)
✓ Story verify:       PASS  (all acceptance criteria met)
✓ Java review:        PASS  (N files checked)
✗ JavaScript review:  FAIL  — formHelper.js: cookie manipulation detected

Action required: Fix 2 issues before proceeding.

Only mark the story done when all applicable subagents show PASS.

---

Subagent Instructions Pattern

When spawning a subagent, provide:

1. The skills folder path so it can read skill-overview.md for the full list and then read any relevant SKILL.md itself
2. The task and scope

Use this prompt structure:

You are a specialized Axon Ivy agent.
Skills folder: .claude/skill/

Start by reading `.claude/skill/axon-ivy/skill-overview.md` to understand all available skills,
then read the SKILL.md file(s) relevant to your task before starting.

Project root: [absolute path]
Task: [what to do]
Scope: [specific files or areas]
Report all issues found or artifacts created.

---

General Rules

• State the detected path in one line before starting — no confirmation needed unless the user explicitly asks
• Show a brief plan (which steps, which subagents) before executing
• On subagent failure, pause the chain — do not skip to the next step
• After any Maven-triggering change (.d.json, pom.xml), remind the user to run mvn install