Skill UI

Condense a long agent conversation at a verified phase boundary by writing a durable checkpoint that preserves constraints, evidence, decisions, artifacts and the next action, so a fresh context can resume the same task without repeating work or losing state.

Runs the Hermes Agent CLI in one-shot mode (hermes -z) to delegate bounded local coding tasks from Code Mate. Covers availability checks, timeouts, exit codes, authentication limits and YOLO tool-approval safety constraints.

Runs Qwen Code headlessly for repository analysis and coding tasks. Delegates work to the AI agent, parses JSON output, and handles authentication errors. Use for automated code review, bug analysis, and refactoring without interactive UI.

This skill covers strategies for compressing conversation history in AI agent sessions to manage context window limits. It focuses on optimizing tokens-per-task rather than just tokens-per-request. Techniques include anchored iterative summarization, opaque compression, and regenerative summaries. It also addresses artifact trail integrity and evaluation methods like probe-based testing to ensure critical information like file paths and decisions are preserved during compression.

Design collaborative AI agent teams with CrewAI: define agent roles, goals and backstories, build dependent tasks, and orchestrate crews via sequential or hierarchical processes with memory, tools, planning and flows.

Automate Datadog monitoring tasks via Rube MCP’s Composio toolkit—query metrics, search logs, manage monitors/dashboards, and create events or downtimes while respecting live tool schemas.

A powerful plugin enabling Claude Code to delegate complex tasks to external AI CLIs (such as OpenAI Codex or Google Antigravity/Gemini) without exiting the conversation. It supports multi-model workflows, allowing users to seek a second opinion, cross-check deep analysis, or delegate side tasks (like research or code review). The core value is that Claude synthesizes the raw outputs, providing a critical review and reconciliation of findings across all consulted models.

The Ejentum Reasoning Harness is a comprehensive library of cognitive operations designed to guide complex AI agents through critical thinking tasks. It exposes four specialized harnesses (reasoning, code, anti-deception, memory) that mitigate common failure modes like hallucination and reasoning decay. Instead of a fixed answer, the agent ingests a detailed, executable scaffold—including failure patterns, procedures, and falsification tests—to ensure robust, self-verified output.

Extract structured fields from documents like invoices, payslips, and contracts with page grounding. Each value cites its source page and confidence score. Missing values are reported as not found instead of hallucinated. Supports PDF, PNG, and DOCX files. Ideal for accounting, compliance, and data parsing tasks where accuracy is critical. Uses Stipple API for processing.

This skill enables AI-powered image editing tasks including style transfer and object removal. It allows users to modify existing images by applying new artistic styles or removing unwanted elements automatically. Suitable for creative workflows requiring image manipulation without manual editing. Ideal for designers and developers integrating AI tools.

Manage AI agent context by offloading data to files for just-in-time loading. Reduces context window bloat and enables dynamic discovery of relevant information. Supports scratch pads, plan persistence, and sub-agent communication to optimize long-term tasks and reduce token costs.

A cross-protocol platform for discovering and searching 18,000+ MCP servers and AI agents across six or more registries. It allows users to find the necessary endpoint for specific tasks (like database access or file conversion) by searching across MCP, A2A, and agents.txt protocols simultaneously.