Skill UI

Execute discrete, bounded coding, review, or verification tasks using the Codex CLI as an isolated subagent. This tool is ideal for tasks requiring explicit file ownership, such as isolated refactoring, feature implementation, or independent code verification, without relying on continuous conversational context. It runs in a safe sandbox environment.

This skill analyzes benchmark run data stored in JSON files to track performance drift over time. It calculates deltas between the first and last runs for metrics like win rate, latency, escalation rate, and LLM-baseline cost. The tool helps identify regressions before release, after corpus expansion, or following agent upgrades by flagging significant changes in speedup or latency compared to baselines.

Build a production-ready, AI-powered data collection agent for any public source (job boards, prices, news, etc.). This agent automatically scrapes data, enriches it using free LLMs (like Gemini Flash), and stores the results in databases such as Notion, Sheets, or Supabase. It runs on a free schedule using GitHub Actions and learns from user feedback, making it ideal for continuous data monitoring and tracking.

This skill runs a single, simulated tick of a defined business pod (Sales). It is designed for dry-run testing, allowing users to validate inputs, resolve agents, and reserve budgeted resources without invoking real LLM workers. It handles template loading, schema validation, budget management, and emits structured status reports for subsequent workflow consumption.

SkillOpt-Sleep enables an AI agent to undergo a simulated "sleep cycle" for continuous, measurable self-improvement. It harvests past sessions, mines recurring tasks, and re-runs them offline. It then consolidates findings into the agent's memory and skills, ensuring only validated improvements are adopted. This process maximizes long-term competence from short-term interactions.

This skill enables the execution of OpenSquilla agents against complex, real-world software engineering benchmarks (SWE-bench). It runs the entire process—from pulling the necessary Docker image to collecting the patch and running official evaluations—in a containerized environment. Use it to test an agent's ability to solve specific coding issues, benchmark performance across datasets (e.g., verified, multilingual), or verify if a generated patch successfully resolves an instance.

Runs a multi-persona dev team session where PM, Architect, Developer, and QA each respond to a topic from their own role. Use for feature design, proposal review, or initiative onboarding when you want multi-role perspectives without switching agents.

This guide details Phase 4: transitioning an interactive agent into a persistent, automated workflow. It covers creating scheduled deployments (cron jobs) or event-driven triggers (cURL). Users learn to validate cron schedules, build the necessary deployment payloads, and follow best practices, including manual testing and DST consideration, ensuring the agent runs reliably without human intervention.

Generate decision-ready reports from AI agent evaluation runs. This skill ensures autonomous, assisted, failed, and timed-out outcomes remain distinct and comparable. It defines metrics with correct denominators, handles latency and cost populations honestly, and maps evidence to decision gates. Ideal for benchmarking, regression testing, and production readiness validation of AI agents.

This skill enables debugging and patching failed Agent QA runs using MCP evidence, artifacts, logs, and local code. It ensures defects are not hidden by rewriting tests. The process involves collecting evidence, classifying failures, inspecting local files, applying minimal changes, and verifying fixes through reruns. It supports self-healing of stale definitions while maintaining safety boundaries.

This skill classifies failed Agent QA runs by analyzing recorded evidence, artifacts, and logs. It determines failure categories, confidence levels, and suggests actionable next steps for debugging. It helps identify issues like timeouts, element not found, or infrastructure problems without guessing, ensuring accurate triage before fixing.

Schedule AI agent runs safely with cron, systemd timers, external clocks, or built-in schedulers such as Hermes, covering heartbeat patterns, permissions, overlap locks, stop conditions and verification to avoid unsafe tight autonomous timers.