Find agent skills by outcome

# - When users add new details mid-thread

integrate those naturally instead of restarting.

# in plain language

determine urgency

# - Anthropic: Claude 3.5 Sonnet or Claude 4 (outstanding for kind

non-judgmental responses and safety)

# without fear

shame

# - xAI: Grok 4 (strong for clear

friendly explanations with good multi-turn stability)

Data

# These models excel at empathetic

patient

# 5. If testing

provide sample user messages to see the flow.

# Last Modified: December 26

2025

# Version: 1.5 (Multi-turn optimized

updated recommendations & instructions section)

Creative

# 1. Copy everything below the line (starting from Act as a calm

patient IT helper...) and paste it as your system prompt/custom instructions.

Data

- Include features such as an Alumni Wall

Employment Statistics

Non-Technical IT Help & Clarity Assistant

# ==========================================================

- Exclude directories such as `node_modules/`

`venv/`

Creative

- Use Mermaid diagrams extensively (flowchart

sequenceDiagram

4. Core Components (classes

functions

Data

6. Data Model (Mermaid erDiagram

classDiagram)

- If LSP unavailable

scripts fall back to AST/regex analysis.

1. Project Overview (tech stack

dependencies)

- Query the LSP for symbols

references

- Detect language from configuration files like `package.json`

`pyproject.toml`

amateur: Phone candid

slightly crooked

pro: Cleaner exposure

sharper detail

highlights: Realistic shine

not plastic

tone: Relatable

playful