Answer: Claude Code evaluates whether a goal is met by sending the condition and the conversation so far to your configured small fast model each time Claude finishes a turn; the model returns one of three verdicts (Not yet met, Met, or Impossible). When a turn fails, Claude Code clears the goal if the error is one you have to fix, and after any other error the goal stays set. Status: Confirmed Entry: tool/claude-code/agents-and-automation/goal-evaluation Revision: 2 Digest: sha256:49983980c0d94eb6 (changes whenever the content changes) Timeline: Created: 2026-11-13T03:04:19Z Last changed: 2026-11-20T12:21:07Z (automatic retries before pause: — → three) Last verified: 2026-11-20T12:21:07Z Next check: 2026-11-21T12:21:07Z Check interval: Daily Expected change: No fixed schedule Data: Field | Value | Source | As of evaluation mechanism | prompt-based Stop hook | [1] | 2026-11-20 evaluator model | small fast model | [1] | 2026-11-20 verdicts | Not yet met | [1] | 2026-11-20 failed turn behavior | Claude Code clears the goal if the error is one you have to fix | [1] | 2026-11-20 warning text | Goal cleared after an unrecoverable error | [1] | 2026-11-20 automatic retries before pause | three | [1] | 2026-11-20 Sources: [1] Anthropic. "Keep Claude working toward a goal". https://code.claude.com/docs/en/goal.md Role: primary. Accessed: 2026-11-20T12:21:07Z Quote: "`/goal` is a wrapper around a session-scoped [prompt-based Stop hook]" (Supports: evaluation mechanism) Quote: "Claude Code sends the condition and the conversation so far to your configured [small fast model]" (Supports: evaluator model) Quote: "**Not yet met**: Claude keeps working and takes the reason as guidance for the next turn." (Supports: verdicts) Quote: "When a turn fails, Claude Code clears the goal if the error is one you have to fix. After any other error the goal stays set." (Supports: failed turn behavior) Quote: "The warning starts with `Goal cleared after an unrecoverable error` and ends with `Run /goal again to continue`." (Supports: warning text) Quote: "After three automatic retries, the goal pauses instead." (Supports: automatic retries before pause) Published: 2026-11-20T02:55:02Z Snapshot SHA-256: dbf77b3a29ea1ea786f5ba35b031274dc4671cd7f264f0f5a5b2fbf0cf1ed78b (captured 2026-11-20T12:21:05Z) Verification: Method: Automated extraction from [1] with deepseek-flash (JSON output, thinking disabled). Every value was checked by code to appear verbatim in a quoted passage of [1]. Cross-check: not-available at 2026-11-20T12:21:07Z. No independent second source is configured for this entry Last successful check: 2026-11-20T12:21:07Z Last attempt: 2026-11-20T12:21:07Z Notes and limitations: Extracted automatically from a single official source (section "How evaluation works"). Related: https://latestn.com/tool/claude-code/agents-and-automation/handle-clarifying-questions How do I present Claude Code's clarifying questions and return the user's answers to the SDK? https://latestn.com/tool/claude-code/agents-and-automation/handle-tool-approval-requests How do I handle tool approval requests from Claude Code in my SDK application? https://latestn.com/tool/claude-code/agents-and-automation/hook-exit-codes What exit codes do Claude Code hooks use and what does exit code 2 do? https://latestn.com/tool/claude-code/agents-and-automation/filter-hook-events How do I filter which events a Claude Code mod hook handles? https://latestn.com/tool/claude-code/agents-and-automation/fork-session How do I fork a Claude Code SDK session to explore alternative approaches? https://latestn.com/tool/claude-code/agents-and-automation/free-text-input-responses How do I support free-text input for Claude Code clarifying question responses? https://latestn.com/tool/ All entries in AI tools Change log: Time | Event | Change | Source 2026-11-20T12:21:07Z | changed | automatic retries before pause: — → three | [1] 2026-11-20T12:21:07Z | changed | failed turn behavior: — → Claude Code clears the goal if the error is one you have to fix | [1] 2026-11-20T12:21:07Z | changed | evaluator model: — → small fast model | [1] 2026-11-20T12:21:07Z | changed | evaluation mechanism: — → prompt-based Stop hook | [1] 2026-11-20T12:21:07Z | changed | other errors: After any other error the goal stays set. → — | [1] 2026-11-20T12:21:07Z | changed | turn failure behavior: Claude Code clears the goal if the error is one you have to fix → — | [1] 2026-11-20T12:21:07Z | changed | verdict impossible: Impossible → — | [1] 2026-11-20T12:21:07Z | changed | verdict met: Met → — | [1] 2026-11-13T03:04:19Z | added | Entry created | — Changes for this entry as a feed: https://latestn.com/tool/claude-code/agents-and-automation/goal-evaluation/latest Challenge this entry: Open challenges: None How: If a value here is wrong or out of date, send what is wrong and a link to evidence. Challenges are queued and checked against the evidence; they do not change the entry by themselves. If the entry changes as a result, the change log says so. Form: https://latestn.com/dispute/tool/claude-code/agents-and-automation/goal-evaluation POST: https://latestn.com/dispute/tool/claude-code/agents-and-automation/goal-evaluation with claim, evidence_url, and optionally field and contact (form-encoded or JSON) GET: https://latestn.com/dispute/tool/claude-code/agents-and-automation/goal-evaluation?claim={what is wrong}&evidence_url={link} How to cite: LatestN. (2026). How does Claude Code evaluate whether a goal is met, and what happens when a turn fails? Revision 2, last verified 2026-11-20T12:21:07Z. https://latestn.com/tool/claude-code/agents-and-automation/goal-evaluation (accessed 2026-11-21T10:01:33Z). @misc{LatestN_tool_claude_code_agents_and_automation_goal_evaluation_r2, title = {How does Claude Code evaluate whether a goal is met, and what happens when a turn fails?}, author = {{LatestN}}, year = {2026}, howpublished = {\url{https://latestn.com/tool/claude-code/agents-and-automation/goal-evaluation}}, note = {Revision 2; digest sha256:49983980c0d94eb6; last verified 2026-11-20T12:21:07Z; accessed 2026-11-21T10:01:33Z} } Reuse: License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) Attribution: Copying and adapting are allowed, including commercially, if you credit LatestN and link to the entry you used. Each entry ends with a ready-made citation. URL: https://latestn.com/tool/claude-code/agents-and-automation/goal-evaluation