Generate images and videos with Agnes AI generation models (agnes-image-2.1-flash, agnes-video-v2.0) via the Agnes apihub REST API. Use when the user asks to generate, create, or draw an image with Agnes, or to generate/create a video with Agnes (text-to-image, image-to-image, text-to-video, image-to-video). Generation only — these models cannot read, describe, or summarize an existing image or video.
Drive interactive terminal programs (REPLs, TUIs, long-running CLIs) through Boo, a screen-style terminal multiplexer with a faithful libghostty terminal emulator. Use this when you need to start an interactive program, send it input, and read its rendered screen deterministically — instead of guessing with sleep or losing output on detach/reattach. Triggers: "run this TUI", "interact with this REPL", "drive an interactive prompt", "program needs a real terminal".
Orchestrate multiple pi subagents with deterministic control flow — the plan lives in code, judgment is delegated to fresh-context subagents. Use when a task is long-running, massively parallel, or adversarial and benefits from isolated subagent contexts: fan-out-and-synthesize (split a task into parallel steps then merge), classify-and-act (route by task type), adversarial verification (a separate agent checks each output against a rubric), generate-and-filter (many candidates, keep the best), tournament (N approaches compete, judged pairwise), loop-until-done (spawn until a stop condition), and deep-research (compose fan-out + verify + synthesize). Mitigates agentic laziness, self-preferential bias, and goal drift.
Web search and content extraction via Brave Search API. FALLBACK web search skill: use tavily-search by default, and use this when Tavily hits its monthly quota/credit/rate limit, or when the user specifically asks for Brave. Also use to search, look up, find, research, check, or query information on the web, fetch/extract content from a URL, find documentation, API references, latest news, current events, or verify facts. No browser needed. Only use web-access instead when interactive page manipulation, login state, or JavaScript execution is required.
Analyzes article structure, identifies positions requiring visual aids, generates illustrations with Type × Style two-dimension approach. Use when user asks to "illustrate article", "add images", "generate images for article", or "为文章配图".
AI image generation with OpenAI, Azure OpenAI, Google, OpenRouter, DashScope, MiniMax, Jimeng, Seedream and Replicate APIs. Supports text-to-image, reference images, aspect ratios, and batch generation from saved prompt files. Sequential by default; use batch parallel generation when the user already has multiple prompts or wants stable multi-image throughput. Use when user asks to generate, create, or draw images.
Clean up local branches whose remote tracking branch is gone. Use when the user says "clean up branches", "delete gone branches", "prune local branches", "clean gone", or wants to remove stale local branches that no longer exist on the remote. Also handles removing associated worktrees for branches that have them.
Commit, push, and open a PR with an adaptive, value-first description. Use when the user says "commit and PR", "push and open a PR", "ship this", "create a PR", "open a pull request", "commit push PR", or wants to go from working changes to an open pull request in one step. Also use when the user says "update the PR description", "refresh the PR description", "freshen the PR", or wants to rewrite an existing PR description. Produces PR descriptions that scale in depth with the complexity of the change, avoiding cookie-cutter templates.