AgentFixboard
UnverifiedLast checked CrewAI#parsing#config

Runtime input prompts write to stdout in front of the run result

prompt_for_inputs writes the field prompt onto stdout. A wrapper that then prints JSON gets topic > \x1b[0m{"topic": "AI"}\n and json.loads fails with Expecting value: line 1 column 3 (char 2) even when the process exits 0.

From GitHub

crewAIInc/crewAI

IssuePull request

Difficulty
Easy
Time
~5 min
Steps
2
Fixed in
—
Last checked

Unverified: we could not confirm from the sources that a fix shipped. Source: PR #8297 merged. Last checked 2026-10-22.

Symptom

prompt_for_inputs writes the field prompt onto stdout. A wrapper that then prints JSON gets topic > \x1b[0m{"topic": "AI"}\n and json.loads fails with Expecting value: line 1 column 3 (char 2) even when the process exits 0.

Evidence what you'll see

  • On Windows with Click 8.4.2 the observed streams were stdout topic > \x1b[0m{"topic": "AI"}\n and stderr \n Crew inputs\n Required values\n.
  • The helper's title, subtitle, and description already use err=True; its click.prompt call did not.
  • Reproduced on main@c6e500ab3730ccf3badb640c6e0170e5c6aa3882 with crewai-cli 1.15.28. JSON crew and declarative-flow input resolution both use this helper.

Why it happens

The field prompt went to stdout, so a consumer could not parse the serialized result. Pull request 8297 passes err=True to Click and, on Windows, redirects stdout only during the prompt because Click 8.4.2 still writes the last prompt character through input() on stdout.

Affected versions

crewai-cli 1.15.28 on an editable main checkout at c6e500ab3730, Click 8.4.2, native Windows, Python 3.12.8. crewAI Tools is not used.

Fix steps

0 of 2 done

Your check marks stay in this browser only.

How to check it worked

  • The same wrapper's stdout is exactly {"topic": "AI"}\n and json.loads(result.stdout) succeeds.

If it still fails

  • No published release tag contains pull request 8297. Tag 1.15.28 was compared against merge 37f0bd32ce and does not contain it.
  • The report says this is the helper's I/O contract, not a claim that every crewai run result is JSON. Existing tests that replace click.prompt do not observe the real streams.

Sources and credit

Summarized in our own words from these public reports. Credit goes to their authors.

Provided as is, without warranty. Try changes in a non-production environment and back up first. Security-related steps: verify against the official docs. Terms · Request removal or attribution change

Hit an error we don't cover? Submit it