# Trial feedback record

[中文](feedback.zh-CN.md) · [Exercise](task.en.md)

This is a blank template, not evidence of an external trial. Record only observations. Mark skipped steps “not run”; a blank is not a pass.

- Anonymous participant code / date:
- Source: external participant / internal rehearsal (choose one)
- Archive version / SHA256 / Python / operating system:
- Data instance / task language / report language:
- Total time / assistance received and what it involved:

| Step | Verified / failed / not run / uncertain | Time | Expected versus actual behavior, error text or evidence file |
|---|---|---|---|
| Create environment and start | | | |
| Explain both task rules | | | |
| Run the approval-request task | | | |
| Run the approved-execution task | | | |
| Select an additional data instance (0.5.0 only) | | | |
| Explain the intentional failure with JSON evidence | | | |
| Read the integration interface | | | |

**Rule understanding:** What determines whether to wait or execute? Which wording is unclear?

**Failure explanation:** What was left incomplete? Were there prohibited attempts or actual violations? Which evidence paths support this?

**Integration obstacles:** What information or interface is missing? How much work would a next step require (unknown is acceptable)? No live model call is needed.

**Usefulness:** Have you encountered a similar error? Which decision would it affect? Would you repeat the same task to test a fix, and why? No purchasing commitment is requested.

**One change that would help most:**

Remove private paths, customer information, tokens and contact details before sharing attachments. The inviter collects feedback manually; this form sends nothing.