lollipopkit

修复, 等待 AI 代码审查评论并处理它们并不断迭代直到可以合并

Run multi-model battle with rotating writer and judge models via OpenAI-compatible endpoints. Use when users ask to battle or compare models, run multi-LLM critique, or iteratively improve an answer across models.

Combine recursive outer-loop refinement with multi-model arena generation each round. Use when users request recursive arena, multi-LLM consensus with iterative refinement, or recursive plus model-battle workflows.

Recursive master-orchestration for complex multi-step tasks. Decompose work, delegate execution to sub-agents, and iterate until verified. Use when users request recursive reasoning, multi-pass refinement, or explicit plan-execute-verify loops.

This skill should be used when the user asks to "analyze a large log", "summarize a big file", "read the whole repository", "process long tool output", "avoid context window overflow", "use sub-agents for preprocessing", "add evidence with line numbers", "verify claims with locators", or mentions "TaskSpec", "SubResult", "Evidence Contract", "Context Firewall", "Map-Reduce", "FileWorker", "Aggregator", "Critic".

Clean Codex local history artifacts that can contain logs, prompts, user input text, command traces, and session transcripts while preserving memories and configuration. Use when the user asks to delete, purge, wipe, prune, reset, or preview Codex history/log/session/input records, especially `~/.codex/history.jsonl`, `~/.codex/sessions/`, `~/.codex/log/`, `~/.codex/logs_*.sqlite`, or shell snapshots, and explicitly not memories.

Create concise, evidence-based handoff notes for another AI coding agent. Use when the user asks for a handoff, continuity note, next-agent summary, progress transfer, "告诉别的 agent", "handoff skill", "交接", or wants the current state, completed work, remaining plan items, risks, and next steps captured for another agent without writing a verbose work log.

Maintain an evolving linked documentation library for project wikis, ADRs, design notes, research notes, release notes, and decision logs. Use when asked to create, grow, repair, reorganize, or verify a wiki/ADR knowledge base with Markdown pages, YAML frontmatter, Obsidian-style [[wikilinks]], source-backed citations, llmwiki/llm-wiki-compiler projects, or a durable project memory that should compound over time instead of being rewritten from scratch.

TrailBase project guide for installing and running the single-executable backend, configuring traildepot/config.textproto and migrations, exposing Record APIs, using auth and token flows, integrating first-party SDKs for TypeScript/JavaScript, Dart/Flutter, Rust, C#/.NET, Swift, Kotlin, Go, and Python, calling HTTP APIs directly, implementing WASM custom endpoints, and locating relevant source/docs in the trailbase repository. Use when Codex needs to answer or implement TrailBase setup, SDK, REST/HTTP API, auth, realtime, configuration, migration, deployment, or custom endpoint work.

Use fl_lib stores, widgets, utils, and extensions in Flutter apps. Use when requests mention fl_lib APIs or integration.

Manage exe.dev virtual machines with the exedev-ctl CLI installed from GitHub Releases. Use when Codex needs to help third-party users install exedev-ctl, inspect, create, delete, restart, rename, tag, resize, SSH into, share, or troubleshoot VMs on exe.dev; generate or check EXE_DEV_API_KEY permissions; or run raw exe.dev commands.

TrailBase project guide for installing and running the single-executable backend, configuring traildepot/config.textproto and migrations, exposing Record APIs, using auth and token flows, integrating first-party SDKs for TypeScript/JavaScript, Dart/Flutter, Rust, C#/.NET, Swift, Kotlin, Go, and Python, calling HTTP APIs directly, implementing WASM custom endpoints, and locating relevant source/docs in the trailbase repository. Use when Codex needs to answer or implement TrailBase setup, SDK, REST/HTTP API, auth, realtime, configuration, migration, deployment, or custom endpoint work.