← ClaudeAtlas

summonerlisted

Routes user intent through summoner.yaml manifest to domain skills. Enforces checkpoint protocol and post-game review. Use when user invokes any /summoner:* command or expresses intent matching a Summoner workflow.
johnson-xue/summoner · ★ 1 · AI & Automation · score 70
Install: claude install-skill johnson-xue/summoner
# Summoner — AI Agent Orchestration Framework ## Overview Summoner is the routing hub. It reads the project's `summoner.yaml` manifest, resolves phase names to domain skills, enforces checkpoint pauses between phases, and triggers post-game review at workflow end. It does NOT implement any domain logic itself — that lives in project skills. ## When to Use - User invokes `/summoner:fix`, `/summoner:new`, `/summoner:ship`, `/summoner:debug`, `/summoner:ops`, or `/summoner:review` - User expresses intent that matches a Summoner workflow (e.g., "帮我排查这个线上 bug" → suggest `/summoner:fix`) **When NOT to use:** - User explicitly says "直接用 my-debug-skill" or names a specific domain skill → route directly, skip Summoner orchestration - Pure Q&A, no workflow needed → respond directly ## Core Operating Behaviors ### 0. Memory Retrieval (Phase 0) Before starting any workflow, check Summoner Memory for relevant historical patterns. 1. Read `project.name` from the project's `summoner.yaml` 2. Check if memory database exists: `memory/{project-name}.db` - If not: run `scripts/init-memory-db.sh {project-name}` to create and seed it 3. Extract features from user input: - error_codes: scan for SC_Err*, "panic", "nil pointer", "index out of range" - module: infer from log file paths (e.g. player/task/task.go → task) - keywords: tokenize Chinese phrases and English words 4. Query patterns table with each extracted feature: ``` SELECT name, type, summary, hits, priority