Text Processing Agent Client Skill

Overview

This skill teaches how to effectively interact with the Text Processing AI Agent's RESTAP endpoints for various text operations.

When to Use This Skill

• Need text analysis or processing
• Want to format or transform text content
• Require document processing capabilities
• Need content validation or cleaning

Agent Interaction Patterns

Basic Text Operations

# Echo text through the agent
curl -X POST http://agent.example.com/text/echo \
  -H "Content-Type: application/json" \
  -d '{"text": "Hello World"}'

Conversational Interaction

# Talk to the agent (one-directional: send query, receive LLM response)
curl -X POST http://agent.example.com/talk \
  -H "Content-Type: application/json" \
  -d '{"message": "How can you help with text processing?"}'

Agent Communication Workflow

1. Discover Capabilities: Check /.well-known/restap.json for available operations
2. Talk First: Use POST /talk endpoint (one-directional: send query, agent receives it and triggers LLM response)
3. Execute Tasks: Call specific capability endpoints based on agent guidance
4. News Endpoint: Use /news as a single bidirectional endpoint:

- GET /news: Read updates (no reply) - POST /news: Write replies/messages (no reply)

Key Points:

• /talk is one-directional - client sends query, agent responds with LLM output
• /news is bidirectional - can read (GET) and write (POST), but never sends a reply

Best Practices

• Always check agent capabilities before making requests
• Use the /talk endpoint to understand proper usage patterns
• Handle both successful responses and error cases
• Respect rate limits and implement appropriate backoff
• Validate response formats before processing

Common Interaction Patterns

• Start with capability discovery via /.well-known/restap.json
• Use /talk for complex requests or when unsure of proper usage
• Implement proper error handling for network issues
• Poll GET /news for asynchronous operation completion
• Use POST /news to send replies directly to other agents (prevents loops)
• Use since parameter when polling: GET /news?since=timestamp
• Cache agent capabilities to reduce discovery overhead

The /news Endpoint: Single Entrypoint for Reading and Writing

The /news endpoint is a single bidirectional entrypoint that handles both reading and writing. The critical property: it never sends a reply.

Reading from /news (GET)

# Poll for updates (no reply triggered)
curl http://agent.example.com/news?since=0

Writing to /news (POST)

# Send reply to another agent (no reply triggered)
curl -X POST http://agent-a.example.com/news \
  -H "Content-Type: application/json" \
  -d '{
    "type": "reply",
    "from": "agent-b",
    "in_reply_to": "query_123",
    "message": "Here is my response..."
  }'

Complete Flow Example

Agent 1 → POST /talk → Agent 2 (one-directional: Agent 2 receives query, triggers LLM response)
Agent 2 → POST /news → Agent 1 (bidirectional write: just stored, no reply)
Agent 3 → GET /news → Agent 2 (bidirectional read: just reads, no reply)

Why this matters:

• /talk is one-directional - client sends query, agent responds with LLM output
• /news is bidirectional but never sends a reply - prevents infinite loops
• When you send a reply via POST /news, the receiving agent doesn't process it - it's just stored
• This allows safe bidirectional communication without triggering endless processing cycles