Skills
Diagnose agent failures by cause: wrong answer → bigger model, abandoned work → higher effort
Anthropic's official guidance separates two knobs that builders routinely conflate. Effort is not thinking time — it governs how many files Claude reads, how many tools it calls, and how many steps it takes before checking back. The diagnostic rule: if Claude had all the context, clearly tried, and was still wrong, raise the model tier; if it skipped a file, never ran the tests, or bailed mid-refactor, raise the effort level. Their stated first instinct for either failure is to fix the context you supplied, not to turn a knob.
↳ Follow the thread