LLM & Agents

Manage PR review feedback loops with Gemini Code Assist. Use when pushing changes to a PR, iterating on review feedback, or monitoring CI/review status. Automatically falls back to Claude when Gemini is rate-limited.

Use when creating or publishing Claude Code hooks - covers executable format, event types, JSON I/O, exit codes, security requirements, and PRPM package structure

Coordinate the recursive improvement loop between Prompt Forge and Skill Forge with a frozen evaluation harness and auditable checkpoints.

Regel 07: Testing. Use when writing tests, reviewing test coverage, or setting up testing.

Run multi-LLM council for adversarial debate and cross-validation. Orchestrates Claude, GPT-4, and Gemini for production-grade implementation, code review, architecture design, research, and security analysis.

React Testing Strategy. Use when writing tests, reviewing test coverage, or setting up testing.

Coordinate distributed agents with resilient topologies, synchronized state, and evidence-backed communication patterns.

This skill should be used when users want to route LLM requests to different AI providers (OpenAI, Grok/xAI, Groq, DeepSeek, OpenRouter) using SwiftOpenAI-CLI. Use this skill when users ask to "use grok", "ask grok", "use groq", "ask deepseek", or any similar request to query a specific LLM provider in agent mode.

Reusable memory patterns (short/long/episodic/semantic) implemented on AgentDB.

Use when testing plans or decisions for blind spots, need adversarial review before launch, validating strategy against worst-case scenarios, building consensus through structured debate, identifying attack vectors or vulnerabilities, user mentions "play devil's advocate", "what could go wrong", "challenge our assumptions", "stress test this", "red team", or when groupthink or confirmation bias may be hiding risks.

Platform selection and orchestration meta-skill across Flow Nexus, Codex, Gemini, and AgentDB.

Push prompt updates to Langfuse (staging or production). Defaults to STAGING (safe). Production requires explicit --production flag + confirmation. NEVER assigns labels (human-in-the-loop safety).

Zero Data Retention mode for sensitive/proprietary code - no code stored on OpenAI servers

DeepSeek AI large language model API via curl. Use this skill for chat completions, reasoning, and code generation with OpenAI-compatible endpoints.

ReasoningBank integrations that rely on AgentDB for memory and retrieval.

AgentDB Reinforcement Learning Training operates on 3 fundamental principles:

Generate and improve prompts using best practices for OpenAI GPT-5 and other LLMs. Apply advanced techniques like chain-of-thought, few-shot prompting, and progressive disclosure.

[CLAUDE CODE ONLY] Leverage Codex CLI for AI peer review, second opinions on architecture and design decisions, cross-validation of implementations, security analysis, and alternative approach generation. Requires terminal access to execute Codex CLI commands. Use when making high-stakes decisions, reviewing complex architecture, or when explicitly requested for a second AI perspective. Must be explicitly invoked using skill syntax.

Telegram automation assistant using telegram-mcp. Use when users want to: (1) Get a digest of unread Telegram messages (2) Analyze their writing style from channel posts (3) Draft and publish posts to Telegram channels (4) Search and reply to messages across chats Triggers: "telegram digest", "unread messages", "morning summary", "post to channel", "draft telegram post", "analyze writing style", "extract style from channel", "telegram workflow"

Use when prompts produce inconsistent or unreliable outputs, need explicit structure and constraints, require safety guardrails or quality checks, involve multi-step reasoning that needs decomposition, need domain expertise encoding, or when user mentions improving prompts, prompt templates, structured prompts, prompt optimization, reliable AI outputs, or prompt patterns.