What problem does it solve? Turning a vague request, GitHub issue, requirements document, or saved plan into a well-structured prompt that a downstream coding agent can execute is hard; this Skill composes problem-describing XML prompts and launches the agent-quorum self-planning loop only after explicit operator confirmation. ## Core Features & Use Cases - Multi-mode input routing: Accepts free-text requests, GitHub issue URLs (fetched via gh issue view), saved requirements documents, or saved plan files, each with a dedicated handling mode. - Pattern-based prompt composition: Matches requests against a 12-pattern registry (architecture audit, code review, security, debugging, refactoring, and more) to select persona, context, and result contract, then emits a 400-700 token XML prompt with role, problem statement, goals, and constraints. - Actuality preflight checks: Verifies that referenced files, docs, exports, CLI flags, and config keys actually exist before embedding claims in the prompt. - Confirmed launch workflow: Saves prompts to .agents/prompts/, presents Thorough/Balanced/Quick run profiles with iteration caps, and starts pnpm run run:cli -- launch detached only after operator approval. - Use Case: Paste a GitHub issue URL, and the Skill fetches the issue, marks it In Progress on the linked Project board, composes a grounded XML prompt, saves it, and offers to launch a detached agent-quorum planning run. ## Quick Start Ask the agent to compose a prompt from your request or issue URL, for example: turn issue https://github.com/org/repo/issues/42 into a prompt and launch a Balanced agent-quorum run.
