An AI agent skill that helps developers decompose complex requests into structured, AI-friendly task sequences. Optimizes context window usage, manages multi-session workflows, and produces better results from fewer, more focused prompts. Compatible with opencode, Claude Code, Cursor, Windsurf, and GitHub Copilot.
- Goal Articulation — clarifies vague requests into precise objectives
- Dependency Mapping — identifies the natural order of work across phases
- Atomic Prompt Slicing — breaks large work into standalone, verifiable prompt cards
- Context Budget Planning — manages context window limits with session handoffs
- Execution Plans — produces ordered, dependency-aware action plans the user can follow
# rolecraft — install from GitHub (recommended)
npx rolecraft install sametcelikbicak/task-decomposer
# or via agentskill.sh
/learn @sametcelikbicak/task-decomposer
# or copy manually
cp -r task-decomposer .opencode/skills/Existing task-decomposition skills focus on agent-to-agent decomposition (AI breaks down work for itself or other agents). This one focuses on human-to-AI interaction — teaching the developer how to structure their prompts and workflow for better AI collaboration.
MIT
