Failure Mode Index · fmi_6_2
Tool-Argument Validity.
Tool Use & Agentic Controlinstrumented · tier 1 · status: stable
The model calls a tool with arguments that violate its schema or references a tool/parameter that does not exist.
The question that exposes it
Call the add(a:number, b:number) tool to sum 10 and 20.
Probes: Whether the tool call's arguments satisfy the schema and the tool exists.
Trap: Models pass strings for numbers, miss required args, or invent a tool.
Signature
Validate each tool call's arguments against the tool schema and the tool registry; failure = schema violation or nonexistent tool/arg.
Example
Calls `get_user(id="all")` where `id` must be an integer.
Severity axes
prevalence
medium
harm
medium
stealth
low
Engineering guidance · class 6
- Validate every tool call against its schema before executing it.
- Cap iterations and detect repeated identical calls. Non-termination is a failure mode, not a hang.
- Diff the final answer against the tool return rather than trusting the summary of it.
- Replay the call log to compute true state. Never ask the model for the running total.
Mitigations
- schema-constrained tool calls
- pre-flight validation
Measurement
No published loop has measured this mode yet. When one does, its cards appear here, newest first.