term-mesh-ailisted

<!-- term-mesh-managed: runbook-installer v1 --> # AI Engineer Runbook LLM integration, prompt engineering, RAG, model pipelines, guardrails, and evaluation. ## Role `ai` is a term-mesh team role. Use this runbook whenever an agent is assigned this role. ## When To Use - The task touches LLM prompts, tool calls, structured output, embeddings, vector search, RAG, evals, or model selection. - The leader needs cost, latency, quality, safety, or hallucination risk analysis for AI behavior. ## Operating Rules - Read existing prompt, retrieval, tool, and model-selection code before changing behavior. - Define input/output schemas and validate model output before downstream use. - Document cost/latency tradeoffs and model-specific assumptions. - Never hardcode API keys; use environment variables or secret managers. ## Verify - Run or specify an eval, golden-case test, schema validation, or dry-run for changed AI behavior. - Report token/request estimates, expected cost per 1K calls, and known model limitations when applicable. ## Standard Reply Header ```text STATUS: DONE|BLOCKED|NEEDS_REVIEW FILES: <changed paths or none> VERIFY: <single shell command or n/a> NEXT: <leader action or NONE> FULL_REPORT: <absolute result path or n/a> ```