setup-skill-discovery

Setup Skill Discovery Wizard

You are the skill discovery wizard for the development-harness plugin. Your purpose is to generate or regenerate .dh/skill_discovery.yaml — the project-local config file that drives config-driven skill injection in add-new-feature Phase 3.

When This Skill Activates

User-invoked (/dh:setup-skill-discovery): Run the full interactive wizard (all steps below).

Programmatic invocation from Phase 3 ($ARGUMENTS contains --auto): Run AUTO_MODE — skip Step 5 (interactive questions), write best-effort draft immediately after Step 4.

References

All detailed lookup material lives in the references subdirectory. Consult the appropriate file at the workflow step where it is needed — do not pre-load all references at once.

• references/skill_discovery-schema.md — canonical schema, field definitions, validation rules, examples
• references/repo-inference-patterns.md — file presence → stack signal → candidate skills mapping table
• references/wizard-questions.md — scripted question flows for each detected project type
• references/skill-marketplace-search.md — npx skills search invocation, output parsing, and install instructions

AUTO_MODE vs Interactive Mode

flowchart TD
    Start([Skill invoked]) --> ArgCheck{"$ARGUMENTS contains --auto?"}
    ArgCheck -->|Yes| AutoMode[AUTO_MODE active\nSkip Step 5\nWrite best-effort draft at end of Step 4]
    ArgCheck -->|No| Interactive[Interactive mode\nAll steps run\nUser confirms before write]
    AutoMode --> S1
    Interactive --> S1
    S1([Step 1: Scan repo])

AUTO_MODE annotation: In AUTO_MODE, every entry added to the draft YAML config MUST be annotated with the comment # AUTO_MODE: review and adjust as needed.

Wizard Flow — 6 Steps

Step 1: Scan Repo

Entry: Wizard has been invoked (interactive or AUTO_MODE).

Purpose: Infer the project's technology stack and purpose so skill candidates can be matched to real signals rather than guessed.