Revert "Require a human-written summary in agent-drafted PR descriptions (#73…" (#73669)

This reverts commit 56bfaf70fe.
This commit is contained in:
Ash Berlin-Taylor authored and GitHub committed 2026-09-24 17:24:38 +03:00
1 parent 654c5e171d
commit 8021e24982
5 files changed
+4 -64

No files matched your search

@@ -62,45 +62,6 @@ For `airflow-core` (and `chart/`, `dev/mypy/`) **user-facing** changes, add a ne
- NEVER add Co-Authored-By with yourself as co-author of the commit. Agents cannot be authors, humans can be, Agents are assistants.
## PR descriptions: Human Summary and AI Summary
An agent-drafted PR description starts with a `## Human Summary`: a few sentences
**typed by the human author in their own words** — why the change is needed, the key
decision or trade-off, and anything reviewers should watch for. The agent-written
description goes below it, collapsed under `## AI Summary`. Reviewers read the Human
Summary as evidence that a person understood and owns the change, so it must never be
produced by an agent. PRs written by humans with the regular template are unaffected.
1. **Ask the user to type the Human Summary** — e.g. "Please type a short Human Summary
(2–5 sentences, your own words: why this change, key decisions, caveats). I will paste
it verbatim." Then wait for their reply.
2. **Paste the reply verbatim.** Do not reword, expand, shorten, translate, reformat, or
"fix" it — typos included.
3. **Never write, draft, suggest, or pre-fill it yourself** — not a proposed wording for
the user to approve, not a paraphrase of something the user said earlier in the
session, not a summary of the commits, not a copy of the AI Summary. If the user asks
you to write it, decline and explain that it has to come from them.
4. If the user prefers to type it in the browser, leave only the `<!-- ... -->`
placeholder under the heading. Do not put any agent text there.
```markdown
closes: #ISSUE (if applicable)
## Human Summary
<!-- The human author's own words, pasted verbatim. Agents must never write this. -->
## AI Summary
<details><summary>Click here</summary>
Agent-written description of what the PR changes and why.
</details>
```
The Gen-AI disclosure block from the repository template follows after a `---`.
## GitHub messages drafted by agents
Anything an agent drafts that ends up posted to GitHub on the user's
@@ -129,30 +129,16 @@ git rebase upstream/<target_branch>
If there are conflicts, resolve them and continue the rebase. If the rebase is too complex,
ask the user for guidance.
Before creating the PR, ask the user to type the `## Human Summary` and paste it
verbatim — never write it yourself. See
[PR descriptions: Human Summary and AI Summary](../airflow-pr-draft-summary/SKILL.md#pr-descriptions-human-summary-and-ai-summary).
Then push the branch to your fork (`origin`) and open the PR creation page in the browser
with the body pre-filled (including the generative AI disclosure already checked):
```bash
git push -u origin <branch-name>
gh pr create --web --title "Short title (under 70 chars)" --body "$(cat <<'EOF'
Brief description of the changes.
closes: #ISSUE (if applicable)
## Human Summary
<!-- The human author's own words, pasted verbatim. Agents must never write this. -->
## AI Summary
<details><summary>Click here</summary>
Agent-written description of what the PR changes and why.
</details>
---
##### Was generative AI tooling used to co-author this PR?
@@ -171,8 +157,7 @@ pre-fills the PR template with the generative AI disclosure already completed.
Remind the user to:
1. Review the PR title — keep it short (under 70 chars), in the imperative mood, and focused on user impact. Do not use Conventional Commits prefixes (`fix:`, `feat:`, `chore:`, …).
2. Check the `## Human Summary` is their own words — or type it in the browser now if they
left it for later.
2. Add a brief description of the changes at the top of the body.
3. Reference related issues when applicable (`closes: #ISSUE` or `related: #ISSUE`).
## Golden rule: when a fix is imminent, open the PR, not an issue
-2
View File
@@ -191,8 +191,6 @@ Before working on an issue, check for existing PRs. Push only to the user's fork
never upstream or `main`. Never list an agent as a commit co-author. Retain the
required Gen-AI disclosure and GitHub-message attribution; do not tag individuals
outside the exceptions in the message guidance. Drafting does not authorize posting.
Agent-drafted PR descriptions start with a `## Human Summary` that the user types and
you paste verbatim — never write, suggest, or paraphrase it yourself.
## apache-magpie framework
-4
View File
@@ -239,10 +239,6 @@ adhere to the following guidelines:
* Make sure to run all relevant static checks and tests locally to verify that the generated code works as
intended and does not introduce any issues.
* State in your PR description that you have used Gen-AI tools to assist in creating the PR.
* When an agent drafts the PR description, start it with a short ``## Human Summary`` that you type
yourself (why the change is needed, key decisions, caveats), and put the agent-written text below
it under a collapsed ``## AI Summary``. Never let the agent write the Human Summary. PRs written
without Gen-AI tools keep using the regular PR template unchanged.
* Be prepared to explain and justify the use of Gen-AI tools if asked by project maintainers or reviewers.
* Remember that the final responsibility for the code in your PR lies with you, regardless of whether
it was generated by a tool or written by you.
+1 -1
View File
@@ -1,3 +1,3 @@
# Generated by dev/skill-evals/eval.py — do not edit or resolve conflicts by hand.
# Run `prek run run-skill-eval --hook-stage manual --all-files` to regenerate.
b845e771c378087ff85e7a813cfcf04659415b1926af932a19857437e4134d31
485f916101818db48c8311bd3e96770695e8681411e9c063a1ff05bc1d051d7c