docs: add a "AI-Generated PR Disclosure" section to AGENTS.md#15758
docs: add a "AI-Generated PR Disclosure" section to AGENTS.md#15758kevinjqliu wants to merge 1 commit intoapache:mainfrom
Conversation
|
Thanks @kevinjqliu for the PR! Should we also talk about AI generated reviews? Maybe add something like "Similarly, if you use AI tools to review PRs, please review the comments for accuracy before posting. Auto-posting AI-generated comments without human review is discouraged." |
| - **Platform/Tool**: The tool or platform that orchestrated the generation (e.g., `GitHub Copilot`, `Cursor`, `Aider`, `Claude Code`, `custom script`). | ||
| - **Human Oversight**: Describe the level of human review applied (e.g., `fully reviewed`, `partially reviewed`, `unreviewed`). | ||
|
|
||
| ### Optional Fields |
There was a problem hiding this comment.
I'm curious if these fields are useful to add. Would this information here help the community? I feel like we could remove these
|
|
||
| ### PR Description Template | ||
|
|
||
| Add the following block at the bottom of AI-generated PR descriptions: |
There was a problem hiding this comment.
We should add this as a section in the PR template so users are prompted to add these if needed.
There was a problem hiding this comment.
I'm not sure how any of this will reliably work _especially filling in model name and version
There was a problem hiding this comment.
Also you could just remove the "AI-generated" here
"Add the following block to the bottom of PR descriptions"
| ``` | ||
| --- | ||
| **AI Disclosure** | ||
| - AI-Generated: Yes |
There was a problem hiding this comment.
This isn't needed, either this section is here or not IMHO
|
|
||
| ## AI-Generated PR Disclosure | ||
|
|
||
| Pull requests that are authored or substantially generated by AI, LLM, or automated agents **must** clearly identify themselves. Include the following information in the PR description: |
There was a problem hiding this comment.
The grammar here is a bit off.
| Pull requests that are authored or substantially generated by AI, LLM, or automated agents **must** clearly identify themselves. Include the following information in the PR description: | |
| Pull requests that are authored or substantially generated by AI, LLM, or automated agents **must** be labeled as "AI-Generated". Include the following information in the PR description: |
There was a problem hiding this comment.
Another full replacement here
When creating a PR description, always include an AI Disclosure block at the bottom. Use the template below and fill in the fields you know (Model, Platform/Tool). For fields you cannot determine (Human Oversight), use [unknown - human to fill in].
Or something like that. Explicit instructions. Minimal branching or ambiguous conditions
| - **Prompt Summary**: A brief description of the prompt or task given to the agent. | ||
| - **Agent Version**: Version of the orchestration tool or agent framework, if applicable. | ||
| - **Iteration Count**: How many agent iterations or attempts were used to produce the final result. | ||
| - **Context Window**: Any notable context limitations that may affect completeness. |
There was a problem hiding this comment.
I'm not sure how 179 or 180 would accurately produced by some agents (or whether that output could be trusted) without an explicit skill or tool to fetch the information
There was a problem hiding this comment.
Needs more instructions less description
|
I generally don't have strong feelings about this in Agent.MD On the con side, everything we put in here is essentially a cost in tokens for all users so I want to make sure it's useful I would recommend we go over the text though and have less "if then" comments and more straight instructions. No need to say "if you are doing this with ai" just say "PR's include this disclosure box after the description" |
|
|
||
| ### Required Fields | ||
|
|
||
| - **AI-Generated**: `Yes` — explicitly state that the PR was generated with AI/LLM assistance. |
There was a problem hiding this comment.
I don't think we need a AI-Generated field, if the disclosure is here we can assume it was
| ### Required Fields | ||
|
|
||
| - **AI-Generated**: `Yes` — explicitly state that the PR was generated with AI/LLM assistance. | ||
| - **Model**: The specific model used (e.g., `Claude Opus 4.6`, `GPT-4o`, `Gemini 2.5 Pro`). Include the version or snapshot date if available. |
There was a problem hiding this comment.
Here would be sometihng like "Set Model to your Model Identifier" or whnot
| - **AI-Generated**: `Yes` — explicitly state that the PR was generated with AI/LLM assistance. | ||
| - **Model**: The specific model used (e.g., `Claude Opus 4.6`, `GPT-4o`, `Gemini 2.5 Pro`). Include the version or snapshot date if available. | ||
| - **Platform/Tool**: The tool or platform that orchestrated the generation (e.g., `GitHub Copilot`, `Cursor`, `Aider`, `Claude Code`, `custom script`). | ||
| - **Human Oversight**: Describe the level of human review applied (e.g., `fully reviewed`, `partially reviewed`, `unreviewed`). |
There was a problem hiding this comment.
Prompt the user to fill this in or leave it as unreviewed
Or something like that
We don't need to describe, we need to instruct
| - **Platform/Tool**: The tool or platform that orchestrated the generation (e.g., `GitHub Copilot`, `Cursor`, `Aider`, `Claude Code`, `custom script`). | ||
| - **Human Oversight**: Describe the level of human review applied (e.g., `fully reviewed`, `partially reviewed`, `unreviewed`). | ||
|
|
||
| ### Optional Fields |
There was a problem hiding this comment.
Probably we don't need to add Optional category. This is for a robot, just tell it to include everything if we think it's important with instructions for missing fields.
|
|
||
| ## AI-Generated PR Disclosure | ||
|
|
||
| Pull requests that are authored or substantially generated by AI, LLM, or automated agents **must** clearly identify themselves. Include the following information in the PR description: |
There was a problem hiding this comment.
Pick one synonym I think. I don't have the context on this, but I can only imagine it would be harder for a model with synonyms rather than just using a single term over and over
|
Why don't we add such a section to the PR template like https://github.com/apache/spark/blob/master/.github/PULL_REQUEST_TEMPLATE? |
5 participants
Follow up to #15529
Add a section for self-disclosure, esp when PRs themselves are created by an agent.
I have seen several newly opened PR that seem to be fully generated; it would be nice to have some kind of indication on the PR description.