Skill UI

This skill provides methods for reliably extracting typed, validated, and structured data from Large Language Models (LLMs), moving beyond simple text parsing. It covers best practices across major APIs (OpenAI's response_format, Anthropic's tool_use, Google's responseSchema) and local models. Key workflows include defining schemas (Pydantic/Zod), implementing constrained decoding, and building robust retry loops for production-grade data pipelines.

Guides founders and GTM teams on designing and deploying AI SDR systems, automating qualification, routing signals into actions, and orchestrating human-in-the-loop handoffs across inbound/outbound motions and CRM stacks.

Autoresearch Agent runs an autonomous optimization loop that edits a target file, invokes a provided evaluation command, keeps metric-improving commits, drops failures, and repeats indefinitely so engineers can optimize speed, bundle size, prompts, or other measurable goals overnight.

Starts a recurring autoresearch experiment loop at user-selected intervals (10 minutes to monthly), auto-schedules CronCreate jobs, records experiment metadata, and offers commands to monitor progress or stop the loop.

This skill orchestrates end-to-end autonomous AI research projects using a sophisticated two-loop architecture. It manages the entire research lifecycle, from initial literature review and hypothesis formation to running constrained experiments and synthesizing findings into publishable papers. It acts as a project manager, autonomously driving the process while structuring the workspace and presenting progress to human users.

This skill automates the scholarly process of refining a generated academic paper. It executes an iterative loop (Review → Fix → Recompile) using advanced LLMs (like GPT-5.5) over multiple rounds. It is designed not for primary research, but specifically for polishing writing quality, correcting theoretical inconsistencies, improving presentation, and enhancing narrative flow after the initial draft is complete. It helps elevate a paper from a draft to a polished submission.

This skill facilitates autonomous, multi-round research review and refinement. It continuously submits work for assessment via external reviewers (Codex or manual), implements suggested fixes, and re-reviews until the designated positive threshold is met or the maximum number of rounds is reached. It manages state persistence and adjusts difficulty levels for rigorous, iterative quality assurance.

This skill automates the research improvement process by creating a closed loop: Review -> Implement Fixes -> Re-review. It leverages any OpenAI-compatible LLM API to iteratively refine research work until a positive assessment is reached or a maximum round limit is hit. Ideal for academic or technical project refinement.

This advanced skill automates the rigorous, iterative review and refinement of complex research topics or papers. It runs multi-round cycles, using an external LLM (MiniMax API) to provide continuous, constructive feedback. The process guides the user through review, fixes, and re-review stages until a positive assessment is achieved or the maximum round limit is reached. It includes robust state persistence for long-running academic workflows.

This solution guides the implementation of a phishing reporting button within email clients, creating a critical feedback loop for security teams. By integrating with SOAR platforms, reported suspicious emails undergo automated triage, including IOC extraction, sandbox analysis, and threat intelligence checks. The workflow enables rapid classification, automated remediation (e.g., mass retraction, sender blocking), and provides detailed feedback to the user reporter, significantly enhancing overall security posture.

This skill automates the entire iterative process of Design Space Exploration (DSE), primarily for computer architecture and EDA problems. It executes complex programs, analyzes resulting objective metrics, and intelligently tunes input parameters (e.g., cache size, pipeline width, optimization passes). It continues iterating until a predefined objective is met or a timeout is reached, making it ideal for optimizing complex system and hardware configurations.

This utility integrates status updates and notifications with Feishu/Lark. It is designed as an internal, zero-impact tool that other complex skills (e.g., experiment runners, review loops) call to report key events like test completion, scoring results, or reaching a checkpoint. It supports both 'push' (fire-and-forget status updates) and 'interactive' (bidirectional approval/rejection) modes, ensuring seamless workflow visibility and robust operation even if configuration is missing.