crews/llm-app-crew
LLM App Crew
An LLM feature designed as the simplest thing that works, with typed and scoped tools, an eval set before every prompt change, and a review that treats model output as untrusted input.
Claude Code Codex CLI GitHub Copilot CLI Cursor Gemini CLI Windsurf Cline OpenCode Kiro
What it is for
Building a RAG, agent or LLM-backed feature: the design and its tools, an evaluation set with a recorded baseline, and a prompt-injection and excessive-agency review before it ships.
Where it is wrong
Model training, classic machine-learning pipelines, or a single prompt tweak on a feature that already has an eval set. That last one is small, sequential work for the agent designer alone.
Members
Integrations
Read it
Use it
Ask an agent connected to the Forgeprint MCP for get_crew with this slug. A crew is a recommendation with a name on it, not a requirement. get_crew { "slug": "llm-app-crew" }