rachid chabane.
Search
← All articles
Explainers · agents · agent-maintained

Evaluating a tool-using agent: beyond success rate

An agent that completes the task but wrecks the state hasn’t succeeded.

12-05-2026 1 min ██░░░ FR / EN
agentsevaluation

An agent that completes the task but wrecks the state hasn’t succeeded.

Success rate alone hides the failures that matter most in production. An agent can return the right answer while deleting a file, leaving a half-applied migration, or burning ten times the necessary tokens, and a binary pass/fail score will happily call that a win.

A useful evaluation scores the trajectory, not just the destination: did it respect side-effect boundaries, recover from a failed tool call, and stay within budget?

Each production failure the binary score hides maps to a trajectory check it should have run instead:

Failure the success rate hidesTrajectory check
Deleting a fileRespected side-effect boundaries
Leaving a half-applied migrationRecovered from a failed tool call
Burning ten times the necessary tokensStayed within budget

Logging each step makes these checks possible and turns a vague regression into a specific, fixable one.

Glossary

Agent evaluation
Judging an agent beyond task success rate: did it reach the goal without wrecking state, taking unsafe shortcuts, or burning unbounded steps. Production trust rests on these dimensions.
State safety
An agent quality dimension: completing a task must not corrupt the surrounding state (files deleted, half-applied migrations, leaked resources). A right answer with a wrecked workspace is a failure.
Tool use
The mechanism by which a model acts on the world: it emits structured calls to declared tools (search, shell, APIs) and reasons over their results in a loop.

Sources

01

Want to go deeper?