Skill UI

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.

This guide details the modern, action-based approach to handling player input in Unity using the Input System package. It covers setting up Input Actions, Action Maps, utilizing the PlayerInput component, and advanced coding patterns (callbacks, polling). Use this when building robust, device-agnostic gameplay controls for modern Unity projects.

This skill allows delegating self-contained coding tasks to assistants like Codex or Claude Code via non-interactive background processes. It supports feature development, code review, and refactoring. Use OpenSquilla's exec_command and process tools to manage sessions, poll output, and send inputs without blocking the main workflow. Ideal for tasks requiring substantial file exploration.

This skill orchestrates coding tasks end-to-end from issue intake to acceptance verification. It operates as a state machine ensuring minimal human intervention by enforcing strict acceptance criteria and verification steps. It handles implementation, review, deployment, and final evidence-backed completion, escalating issues when DoD fails or permissions are missing.

Audits GitHub Actions workflows that invoke AI coding agents (Claude Code Action, Gemini CLI, OpenAI Codex, GitHub AI Inference). Detects attack vectors where attacker-controlled input reaches AI agents in CI/CD pipelines, covering trigger events, env var intermediaries, tool allowlists and sandbox configuration.

This skill creates, revises, or audits AGENTS.md files based on repository evidence and verified commands. It ensures coding agents understand the workflow without rediscovering it, preserving maintainer intent. It builds evidence maps from manifests and CI, scopes instructions correctly, and validates changes before handoff to prevent mistakes.

Audit local AI coding-agent sessions with agenttrace to inspect token and cost spikes, tool failures, retry loops, latency, anomalies, health scores, diffs, and CI gates. Use for debugging slow, expensive, or unreliable runs and comparing attempts.

This skill allows you to delegate bounded coding tasks to the Google Antigravity CLI (agy) while retaining review and landing responsibility. You write the brief, dispatch the task, wait for completion, review the generated diff against gates, and commit the changes yourself. It requires the agy CLI installed and authenticated, suitable for orchestrators managing code generation workflows.

Secure AI coding agents with permission boundaries, secret protection, code review gates, and safe sandbox configurations for team environments. Covers Claude Code, Cursor, Copilot, and Codex security implementations including CLAUDE.md configuration, command allowlists, filesystem access controls, and pre-commit hooks for secret scanning.

This toolkit provides six expert-level workflows for AI engineering, including prompt evaluation, context budget planning, RAG pipeline design, agent security audits, evaluation harness building, and product sense coaching. It transforms AI coding assistants into senior partners by offering structured methodologies, quantitative scoring, and decision frameworks. Suitable for optimizing LLM systems, designing secure agents, and planning product strategy before development.

Orchestrates bounded coding tasks to Aider CLI, managing briefs, diffs, and reviews. Supports local models via OpenAI-compatible endpoints, ensuring offline operation and no auto-commits.

Forensic analyzer for AI coding sessions. Identifies root causes of scope drift and rework by classifying user specs, agent errors, and repo friction. Generates evidence-backed reports to improve future prompts and project health.