Skill UI

RecallMax dramatically enhances AI agent memory capabilities by cleanly injecting massive external contexts (500K-1M tokens) without hallucination drift. It automatically summarizes lengthy conversations, preserving tone, intent, and key facts. Furthermore, it compresses multi-turn interaction histories into high-density token sequences, ensuring agents maintain perfect recall across complex and long-running sessions.

This skill implements advanced 'Gatekeeper' logic to optimize AI agent performance in long-running workflows. It prevents context window bloat by sharding metadata, allocating token budgets, and enforcing Atomic Precision to deliver purely functional, filler-free responses. Ideal for complex coding, multi-step reasoning, and large document analysis where conciseness is critical.

Create repository-based execution contracts for multi-session implementation work, with stable task IDs, evidence gates, checkpoints, and exact resume state for fresh agents.

Review supervisor, swarm, planner-worker and parallel agent workflows as distributed state machines: task ownership, shared state, branch joins, retries, cancellation, context handoffs, budgets, deadlocks and human escalation before rollout.

This skill enables parallel read-only multi-agent review of git diffs or specific file scopes. It identifies behavioral regressions, security risks, performance issues, and coverage gaps using four specialized sub-agents. The main agent synthesizes findings to filter and summarize critical issues without modifying code. Ideal for ensuring code quality and safety before merging changes.

Coordinate multi-layer security scanning and hardening across application, infrastructure, and compliance controls using defense-in-depth strategy. Orchestrates specialized security agents for comprehensive assessments, layered controls, and continuous monitoring. Follows DevSecOps principles with shift-left security, automated scanning, and compliance validation to create a resilient security posture addressing vulnerabilities and future threats.

This skill scaffolds the necessary configurations for advanced engineering workflows. It guides the user through setting up the issue tracking system (GitHub, GitLab, or local markdown), defining the canonical triage label vocabulary, and establishing the domain documentation context (single or multi-context). It must be run once to initialize the repository for subsequent agent-driven development skills.

This skill reviews code differences (diffs) to identify opportunities for improvement in clarity, reusability, efficiency, and overall quality. It employs multiple specialized sub-agents (e.g., Code Reuse, Efficiency, Quality) to audit changes. It can provide high-confidence, behavior-preserving fixes when requested, ensuring the code adheres to best practices without introducing regressions. Best used when refactoring, cleaning up, or auditing a specific code scope.

Freezes PRD, technical design, and acceptance criteria before medium-to-large Codex work, then coordinates multiple agents with explicit file ownership and judges delivery from diffs, tests, and evidence-backed acceptance.

A quota-aware parallel subagent coordination skill for Antigravity 2.0. Decomposes large multi-file tasks into isolated agent missions, manages model selection based on cost, and ensures efficient parallel execution while monitoring quota usage. Ideal for complex projects spanning multiple files and components.

Expert TDD orchestrator for enforcing red-green-refactor discipline, coordinating multi-agent workflows, and applying modern test-driven development practices across software projects. Covers test suite architecture, AI-assisted test generation, metrics, and framework integration.

An advanced reliability wrapper for AI agents that monitors, manages, and recovers from tool call failures. It automatically handles common issues such as truncated JSON responses, API timeouts, rate limits, and mid-chain process breakdowns. By implementing sophisticated failure classification, checkpointing, and learning mechanisms, Guardian ensures complex, multi-step agent workflows remain stable and resilient.