Drill · lap 1 · 2/9 · AGL-102 · Two bugs in a working dispatcher

[akash-coded]

2 of 9 · AGL-102 · Two bugs in a working dispatcher

agent-loop · medium · fix · ~12 min · no AWS account

This dispatcher passes its unit tests. It has run in production. It has two bugs, and both have ended real runs.

def dispatch(tool_use, registry):
    tid = tool_use.get("toolUseId")
    name = tool_use.get("name")
    args = tool_use.get("input") or {}

    fn = registry.get(name)
    if fn is None:
        return None                                   # unknown tool

    try:
        result = fn(**args)
    except Exception as exc:
        return {"toolResult": {"toolUseId": tid, "status": "error",
                               "content": [{"text": f"{type(exc).__name__}: {exc}"}]}}

    return {"toolResult": {"toolUseId": tid, "status": "success",
                           "content": [{"json": result}]}}

Find them

Both bugs are in what the function lets escape — one is a value that escapes, one is an exception that does. Read the code as the loop that calls it: what happens next, on the very next turn, in each case?

Fix the code so that every path returns a well-formed toolResult block, and the unknown-tool message names the tools that do exist.

Post the fixed function:

/drill AGL-102

```python
def dispatch(tool_use, registry):
    ...
```

What this proves

That you think about a dispatcher as a boundary — the one place whose entire job is to convert anything at all into a value the loop can continue from. The two bugs here are the two most common ways a boundary leaks.

---

Submit in this thread

Copy the block from the brief, edit it, and post it as a comment here. The bot replies within a couple of minutes with every check, the skill you demonstrated, and what to do next.

_{Brief in the repo: labs/drills/agent-loop/AGL-102 · next in the sequence: TOOL-101 · how grading works · scoreboard}