What problem does it solve?
Manually editing JSON configuration files to tune how prd-taskmaster behaves is error-prone and tedious. This Skill replaces raw JSON editing with a guided interview: the AI asks curated questions in plain English, validates each answer, and writes the result to .atlas-ai/config/atlas.json so future runs apply your preferences automatically.
Core Features & Use Cases
- Guided configuration interview: Asks one curated question at a time covering provider, validation strictness, execution mode, template, autonomous behavior, and phase gates, then maps plain-English answers to config keys.
- Validated, idempotent writes: Every answer is validated via
script.py validate-answer before being written, and re-running the skill cleanly updates the existing per-project config.
- Escape hatch for power users: Settings beyond the curated questions live in
.atlas-ai/customizations/, which the skill never overwrites.
- Use Case: You want prd-taskmaster to default to Gemini for task generation, use lenient PRD validation, and prefer execution mode B. Say "customise workflow", answer three questions, and the config is written and verified.
Quick Start
Ask the AI to customise the prd-taskmaster workflow and answer the curated questions about provider, validation strictness, and execution mode.