dx-adapt

You detect the project's technical profile from source files and update .ai/config.yaml with the project:, toolchain:, and additional build: fields. Then you substitute the detected values into .claude/rules/ files (replacing generic placeholders with real project values).

Use ultrathink — project detection requires reading multiple source files and making accurate inferences.

0. Check Prerequisites

Read .ai/config.yaml. If it doesn't exist: "Run /dx-init first." STOP.

Check if .ai/config.yaml already has a project: section with type field:

• If no project: section: Go directly to Phase 1. (This is normal on first run or when called from dx-init.)
• If project: section exists: Display the current values and ask: (A) Keep as-is, (B) Re-detect (re-reads source files), (C) Edit (re-confirm with current values pre-loaded). If A: skip to Phase 4 (rule substitution only).

Phase 1: Detect Project Type

1a. Auto-detect (or use argument)

If the user provided an argument (aem-fullstack, aem-frontend, frontend), use it. Otherwise detect:

Check	Type
pom.xml exists AND contains com.adobe.cq or AEM uber-jar dependency	aem-fullstack
pom.xml exists AND contains content-package-maven-plugin AND no Java source in src/main/java/	aem-frontend
package.json exists AND contains aem-clientlib-generator or clientlib.config.js exists	aem-frontend
package.json exists AND no AEM indicators	frontend
Only pom.xml (no AEM)	java

1b. Extract details per project type

For aem-fullstack:

# Find Java package root
find . -path "*/src/main/java/*.java" | head -1
# → extract package from first line of that file

# Find component namespace from filter.xml or .content.xml
grep -r "componentGroup" --include="*.xml" -l | head -5

# Find base branch
git symbolic-ref refs/remotes/origin/HEAD 2>/dev/null | sed 's|refs/remotes/origin/||'

# Find Maven modules
find . -maxdepth 3 -name "pom.xml" | grep -v "^./pom.xml"