Skip to content
Back to skills

1097 Troubleshooting Ad1270b4

CSecurity

This document provides diagnostic procedures and resolution strategies for common issues encountered during Promptheus operation.

  • 9 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 11, 2026
toolspythongoshellbashdebugginggitapiperformance

Works with

  • terminal
  • cli
  • api
  • mcp

Security analysis

C71/100
  • criticalModifies startup scripts or system services for persistence
  • criticalModifies startup scripts or system services for persistence
  • mediumInstalls packages at runtime which could introduce malicious dependencies

Pro shows the line behind each finding and how to fix it

Scanned October 11, 2026

npx -y skills add tools-only/X-Skills --skill 1097-troubleshooting_ad1270b4 --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of 1097 Troubleshooting Ad1270b4?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for 1097 Troubleshooting Ad1270b4
[![Security: C — Skills Directory](https://www.skillsdirectory.com/api/skills/tools-only-1097-troubleshooting-ad1270b4/badge)](https://www.skillsdirectory.com/skills/tools-only-1097-troubleshooting-ad1270b4)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

SKILL.md
# Promptheus Troubleshooting Guide

This document provides diagnostic procedures and resolution strategies for common issues encountered during Promptheus operation.

## Provider Configuration Issues

### Provider Auto-Detection Failure

**Symptom:** System fails to detect configured providers

**Resolution:**
```bash
# Explicitly specify provider
promptheus --provider google "Prompt"
promptheus --provider anthropic "Prompt"
promptheus --provider openai "Prompt"
promptheus --provider groq "Prompt"
promptheus --provider qwen "Prompt"
promptheus --provider glm "Prompt"
```

### Incorrect Provider Selection

**Symptom:** System selects unexpected provider

**Diagnostic Steps:**
```bash
# Verify environment configuration
cat .env

# Check environment variables
env | grep -E '(GOOGLE|ANTHROPIC|OPENAI|GROQ|DASHSCOPE|ZHIPUAI|PROMPTHEUS)'

# Override provider selection
export PROMPTHEUS_PROVIDER=google
```

### API Key Validation

**Verification Commands:**
```bash
# Validate specific provider configuration
promptheus validate --providers google
promptheus validate --providers anthropic
promptheus validate --providers openai
promptheus validate --providers groq
promptheus validate --providers qwen
promptheus validate --providers glm

# Test live API connectivity
promptheus validate --test-connection
```

## Installation and Runtime Issues

### Command Not Found Error

**Symptom:** Shell reports `promptheus` command not found

**Resolution:**
```bash
# Reinstall in editable mode
pip install -e .

# Execute via Python module
python -m promptheus.main "Prompt"

# Verify installation
which promptheus
```

### Dependency Resolution Errors

**Symptom:** Import errors or missing module exceptions

**Resolution:**
```bash
# Install required dependencies
pip install -r requirements.txt

# Force reinstallation
pip install -e . --force-reinstall
```

## File Input Issues

### File Not Found Errors

**Symptom:** File input fails with path resolution errors

**Resolution:**
```bash
# Use absolute path
promptheus -f /absolute/path/to/prompt.txt

# Use relative path from current directory
promptheus -f ./relative/path/prompt.txt
```

## Clipboard Integration Issues

### Platform-Specific Requirements

**Linux:**
- Install `xclip` or `xsel`: `sudo apt-get install xclip`

**macOS:**
- Native clipboard support (no additional installation required)

**Windows:**
- Native clipboard support (no additional installation required)

**WSL (Windows Subsystem for Linux):**
- May require X server configuration for clipboard access
- Verify X server is running and DISPLAY variable is set

## Interactive Mode Issues

### Multiline Input Problems

**Symptom:** Unable to insert newlines in prompt input

**Resolution:**
- Use `Shift+Enter` for newline insertion
- Alternative: `Option/Alt+Enter` or `Ctrl+J`
- Fallback: Use plain mode if terminal is unresponsive

### Slash Command Failures

**Common Issues:**
- Commands must begin with `/` prefix (not `:`)
- Use `/help` to display available commands
- Use Tab key for command completion suggestions

### Session Configuration Management

**Commands:**
```bash
# Display current session configuration
/status

# Modify provider
/set provider google

# Modify model
/set model gpt-4o

# Toggle modes
/toggle refine
/toggle skip-questions
```

## Shell Completion Issues

### Completion Script Not Functional

**Symptom:** Tab completion does not work after installation

**Diagnostic Steps:**
```bash
# Verify completion script generation
promptheus completion bash  # Should output script

# Verify promptheus is in PATH
which promptheus

# Reload shell configuration
source ~/.bashrc   # Bash
source ~/.zshrc    # Zsh

# For Poetry environments, create alias
alias promptheus='poetry run promptheus'
```

## Model Discovery Issues

### Model List Not Updating

**Symptom:** Model lists appear outdated or do not include newly available models

**Resolution:**
```bash
# Manually refresh model cache
promptheus validate --test-connection  # This will also refresh cache

# Or, in Web UI:
# Go to Settings and click "Refresh Model Cache" button
```

**Cache Location:**
- On Unix: `~/.promptheus/models_cache.json`
- On Windows: `%APPDATA%/promptheus/models_cache.json`

## Reporting Issues

When reporting issues, include the following information:

1. **Command executed:** Full command with all flags
2. **Error output:** Complete error message and stack trace
3. **Environment details:**
   - Python version (`python --version`)
   - Operating system and version
   - Provider and model in use
   - Installation method (pip, source, Poetry)
4. **Reproduction steps:** Minimal steps to reproduce the issue
5. **Current model cache status:** If issue related to model discovery

**Issue Tracker:** https://github.com/abhichandra21/Promptheus/issues

## MCP Server Issues

### MCP Package Not Installed

**Symptom:** Error message when starting MCP server
```
Error: The 'mcp' package is not installed. Please install it with 'pip install mcp'.
```

**Resolution:**
```bash
# Install MCP package
pip install mcp

# Or install with development dependencies
pip install -e .[dev]

# Verify installation
python -c "import mcp; print('MCP installed successfully')"
```

### MCP Server Connection Issues

**Symptom:** MCP client cannot connect to Promptheus MCP server

**Diagnostic Steps:**
```bash
# Verify MCP server starts correctly
promptheus mcp

# Check for error messages in server output
# Test with direct Python execution
python -m promptheus.mcp_server

# Verify provider configuration
promptheus validate --test-connection
```

**Common Causes:**
- Missing provider API keys
- MCP package not properly installed
- Port conflicts (MCP uses stdio transport)

### MCP Tool Execution Errors

**Symptom:** MCP tools return error responses

**Diagnostic Commands:**
```bash
# Test individual MCP tools
# Use MCP client to call list_providers
# Use MCP client to call validate_environment

# Check provider status
promptheus list-providers

# Validate environment configuration
promptheus validate --providers google,openai
```

**Common Error Types:**
```json
{
  "type": "error",
  "error_type": "ConfigurationError",
  "message": "No provider configured. Please set API keys in environment."
}
```

**Resolution:**
```bash
# Configure provider API keys
promptheus auth google
promptheus auth openai

# Or manually set in .env file
echo "GOOGLE_API_KEY=your_key_here" >> .env
echo "OPENAI_API_KEY=your_key_here" >> .env
```

### AskUserQuestion Integration Issues

**Symptom:** MCP server not handling clarification questions properly

**Understanding the Integration Modes:**

**Interactive Mode** (when AskUserQuestion is available):
- Server automatically asks questions via injected AskUserQuestion function
- Seamless user experience within supported clients

**Structured Mode** (fallback):
- Server returns `clarification_needed` response with formatted questions
- Client responsible for calling AskUserQuestion tool
- Answers mapped back via `answer_mapping` dictionary

**Debugging AskUserQuestion Issues:**
```bash
# Test refine_prompt workflow
# Step 1: Call with basic prompt that should trigger questions
# Step 2: Verify clarification_needed response format
# Step 3: Check questions_for_ask_user_question array format
# Step 4: Verify answer_mapping structure
```

**Expected Response Format:**
```json
{
  "type": "clarification_needed",
  "questions_for_ask_user_question": [
    {
      "question": "Who is your target audience?",
      "header": "Q1",
      "multiSelect": false,
      "options": [
        {"label": "Technical professionals", "description": "Technical professionals"}
      ]
    }
  ],
  "answer_mapping": {
    "q0": "Who is your target audience?"
  }
}
```

### MCP Client Compatibility Issues

**Symptom:** MCP client cannot properly communicate with Promptheus MCP server

**Resolution:**
```bash
# Ensure MCP client supports:
# - Tool calling capabilities
# - JSON response processing
# - AskUserQuestion tool (for interactive mode)

# Test with basic tool calls first
# Verify response format compatibility
# Check error handling consistency
```

### Performance Issues with MCP Server

**Symptom:** Slow response times from MCP tools

**Diagnostic Steps:**
```bash
# Test provider connectivity
promptheus validate --test-connection

# Check for model fallback behavior
# Monitor API response times
# Verify cache utilization (model cache, etc.)
```

**Optimization Tips:**
- Use specific provider/model parameters to avoid auto-detection overhead
- Ensure provider API keys are valid to prevent retry delays
- Consider using cached model information when available

### MCP Server Logging and Debugging

**Enable Debug Logging:**
```bash
# Set logging level for MCP server
export LOG_LEVEL=DEBUG

# Run MCP server with verbose output
promptheus mcp --verbose  # If supported by client
```

**Common Debug Information:**
- Provider initialization sequence
- Tool execution timing
- Question generation process
- Answer mapping validation

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.