super-llm-wikilisted

# Super LLM Wiki **🦊 阿舟 · Super LLM Wiki** > 📚 知识要留得住，也要经得起查证。 Use a project-local Markdown knowledge base for verified architecture, decisions, patterns, debugging facts, environment notes, conventions, references, and reviewed session learnings. ## Brand and progress For an interactive run, read [brand-layer.md](references/brand-layer.md), then start exactly once: ```text 🦊 阿舟 · Super LLM Wiki 启动｜operation=<operation>｜scope=<project-root> ``` Use one fixed anchor per completed material stage. Keep machine JSON, paths, commands, schema values, page content, and raw evidence emoji-free. Never emit `✅ 验证通过` after a `fail`, `hold`, or `skipped` machine receipt. A host without Unicode may remove emoji without changing prefixes, separators, fields, or values. Every runtime entry uses the canonical `<project>/.azhou/super-llm-wiki/` store. The CLI is `scripts/llm_wiki.py`; the optional stdio MCP server is `scripts/llm_wiki_mcp.py`. ## Operating contract 1. Resolve the project root before reading or writing. Normal operations never select an alternate store. 2. Query or list before adding related knowledge. Use `ingest` to append a sourced update; use `add` only when duplicate titles should fail. 3. Record evidence in `--source`, choose an honest confidence, and exclude secrets, raw private transcripts, tokens, and unrelated personal data. 4. Run `lint --no-log` after mutations. Broken references, invalid pages, and `implemented`/`rejected` decision pages without an `#