drin-prompt-chain-debugger

Prompt Chain Debugger

Purpose

To ensure composite AI systems remain reliable and behave according to set specifications.

Debugging Steps

1. Trace: Map the flow from input through every agent/prompt node.
2. Inspect Output: Check the raw string output of every node.
3. Validate Interface: Did Node A provide what Node B expected?
4. Sensitivity Test: Does a small change in input drastically break the chain?

Rules

• Log the temperature and model used for every step of the chain.
• If a chain is fragile, break it down into more discrete, verifiable steps.

---

subagents:

• prompt-engineer
• debugger