Why intent drift matters
The most expensive agent failures are not always syntax errors. They are plausible completions that miss the point while sounding finished.
An agent may add a fallback instead of fixing the root cause, change a different path than the requested one, or implement a narrower behavior than the acceptance criteria require.
How FeelGoot maps intent
FeelGoot starts from the user request, acceptance criteria, constraints, and non-goals. It then checks whether changed files and tests are connected to those requirements.
The output is not merely a score. It is a reviewable map of what is supported, what is partial, and what does not connect to the task.
Common drift patterns
Adjacent feature: the change improves a related workflow but not the requested one.
Narrow patch: the code handles a single example but not the general behavior.
Summary inflation: the final answer claims complete work even when evidence supports only partial completion.
Context loss: the agent ignores constraints, non-goals, or prior decisions.