Answer: Have the coding agent call tools directly and fix any that return the wrong evidence before it touches the instructions, then run one real turn, read the trace that `run` prints, and run the smoke evals. The commands include `agent-sdk validate --dir .`, `agent-sdk info --dir . --json`, `agent-sdk tool call inspect_pr --dir . --input '{"prUrl":"https://github.com/acme/checkout/pull/42"}'`, `agent-sdk run --dir . --message "Is https://github.com/acme/checkout/pull/42 ready to approve?"`, `agent-sdk session get --file `, and `agent-sdk eval --dir . --tag smoke`. Status: Confirmed Entry: tool/cursor/troubleshooting/verify-tools-before-tuning-prompt Revision: 1 Digest: sha256:604d4c9e84dcddfa (changes whenever the content changes) Timeline: Created: 2026-11-13T03:08:24Z Last changed: 2026-11-13T03:08:24Z Last verified: 2026-11-20T12:32:34Z Next check: 2026-11-21T12:32:34Z Check interval: Daily Expected change: No fixed schedule Data: Field | Value | Source | As of validate command | agent-sdk validate --dir . | [1] | 2026-11-20 info command | agent-sdk info --dir . --json | [1] | 2026-11-20 tool call command | agent-sdk tool call inspect_pr --dir . --input '{"prUrl":"https://github.com/acme/checkout/pull/42"}' | [1] | 2026-11-20 run command | agent-sdk run --dir . --message "Is https://github.com/acme/checkout/pull/42 ready to approve?" | [1] | 2026-11-20 session get command | agent-sdk session get --file | [1] | 2026-11-20 eval command | agent-sdk eval --dir . --tag smoke | [1] | 2026-11-20 Sources: [1] Anysphere. "cursor.com/docs/agent-sdk/building-with-agents.md". https://cursor.com/docs/agent-sdk/building-with-agents.md Role: primary. Accessed: 2026-11-20T12:32:34Z Quote: "agent-sdk validate --dir ." (Supports: validate command) Quote: "agent-sdk info --dir . --json" (Supports: info command) Quote: "agent-sdk tool call inspect_pr --dir . --input '{"prUrl":"https://github.com/acme/checkout/pull/42"}'" (Supports: tool call command) Quote: "agent-sdk run --dir . --message "Is https://github.com/acme/checkout/pull/42 ready to approve?"" (Supports: run command) Quote: "agent-sdk session get --file " (Supports: session get command) Quote: "agent-sdk eval --dir . --tag smoke" (Supports: eval command) Snapshot SHA-256: e23539ac6cf0ede57a1ab6e12f1de6a330f3620b9bff1f089b07502b9cdd9b89 (captured 2026-11-13T03:08:24Z) 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]. On this check the text of [1] was unchanged since 2026-11-13T03:08:24Z, so the values verified then were kept. Cross-check: not-available at 2026-11-20T12:32:34Z. No independent second source is configured for this entry Last successful check: 2026-11-20T12:32:34Z Last attempt: 2026-11-20T12:32:34Z Notes and limitations: Extracted automatically from a single official source (section "Verify tools before tuning the prompt"). Related: https://latestn.com/tool/cursor/troubleshooting/why-endpoint-exclusions-needed Why does Cursor need endpoint security exclusions and what happens if they are missing? https://latestn.com/tool/cursor/troubleshooting/windows-endpoint-security-exclusions Which Cursor processes and paths should I add to my endpoint security exclusion list on Windows? https://latestn.com/tool/cursor/troubleshooting/worker-offline-live-workspace Why does the Cursor live workspace say Worker offline? https://latestn.com/tool/cursor/troubleshooting/tour-too-large Why does Cursor show 'Too large for a tour' on a pull request? https://latestn.com/tool/cursor/troubleshooting/troubleshoot-xcode-tools-server How do I troubleshoot Cursor not finding the xcode-tools server? https://latestn.com/tool/cursor/troubleshooting/verify-endpoint-exclusions How do I verify Cursor exclusions are working and troubleshoot if they are not? https://latestn.com/tool/ All entries in AI tools Change log: Time | Event | Change | Source 2026-11-13T03:08:24Z | added | Entry created | — Changes for this entry as a feed: https://latestn.com/tool/cursor/troubleshooting/verify-tools-before-tuning-prompt/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/cursor/troubleshooting/verify-tools-before-tuning-prompt POST: https://latestn.com/dispute/tool/cursor/troubleshooting/verify-tools-before-tuning-prompt with claim, evidence_url, and optionally field and contact (form-encoded or JSON) GET: https://latestn.com/dispute/tool/cursor/troubleshooting/verify-tools-before-tuning-prompt?claim={what is wrong}&evidence_url={link} How to cite: LatestN. (2026). How do I verify Cursor Agent SDK tools before tuning the agent prompt? Revision 1, last verified 2026-11-20T12:32:34Z. https://latestn.com/tool/cursor/troubleshooting/verify-tools-before-tuning-prompt (accessed 2026-11-21T10:03:34Z). @misc{LatestN_tool_cursor_troubleshooting_verify_tools_before_tuning_prompt_r1, title = {How do I verify Cursor Agent SDK tools before tuning the agent prompt?}, author = {{LatestN}}, year = {2026}, howpublished = {\url{https://latestn.com/tool/cursor/troubleshooting/verify-tools-before-tuning-prompt}}, note = {Revision 1; digest sha256:604d4c9e84dcddfa; last verified 2026-11-20T12:32:34Z; accessed 2026-11-21T10:03:34Z} } 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/cursor/troubleshooting/verify-tools-before-tuning-prompt