---
name: pushing-commits-to-the-repo
description: Open and advance a PR — write a current title and body, label it, review every push
  subject to the mechanical-refresh exception, watch CI, and triage every comment. Use whenever you
  open a PR or push a commit to one.
---

# pushing-commits-to-the-repo

Pushing starts a loop; it does not end the task. **Work stops only when CI is green AND no comment
is left unresolved.**

## When you open the PR

### Write the title and body

Follow the title and template rules in the root `AGENTS.md`.

Keep visible body content within 40 lines. Exclude template lines and collapsed `<details>`
contents from the count. For a feature or behavior change, use this order:

1. **Why we make these changes** — State the problem and decision in a few sentences. Link the issue.
2. **New public surface** — List each new maintained symbol. Write `none` when there is none.
3. **User-visible behavior** — Show the smallest before-and-after example. Replace it with a
   call-path diff when the changed call chain explains the behavior; do not include both.
4. **Verification** — Link the exact proving tests from the PR diff. Put a minimal runnable
   playground in `<details>` only when it helps reviewers reproduce the behavior.
5. **What changes for existing users** — State the effect in one sentence. `Nothing` is valid.

Use one collapsed `<details>` section per goal only when the PR has multiple independent goals.
For a trivial PR, use the issue link, a short summary, and the test plan.

#### User-visible call-path diff

Use one fenced `diff` tree from the public entry point to the changed observable result.

- Format each node as `path/file.py :: Class.method()` or `path/file.py :: function()`.
- Indent each callee beneath its caller with `└─`. Preserve enough unchanged nodes to show each edge.
- Collapse irrelevant intermediate calls as `… unchanged machinery …`.
- Include arguments only when they explain the change.
- Include results only on relevant leaves.
- Keep the shared caller prefix unmarked. Mark only diverging nodes, relevant arguments, or results.
- Target 12 content lines inside the fence. Never exceed 20; collapse secondary branches instead.

### Keep labels current with the title and body

Assign title, body, and label updates to the same subagent. Reconcile labels whenever the title or body changes.

Run `.agents/skills/pushing-commits-to-the-repo/label-catalog` and give the subagent the printed file path.
The checked-in catalog supplies clones and worktrees without label-read requests.
Run the helper with `--refresh` after creating, renaming, or updating repository labels.
Commit refreshed catalog changes.
The helper uses at most two requests and verifies completeness before replacing the catalog.
Never fetch labels individually or pass raw label API responses.

Keep the category label aligned with the PR's purpose (`bug`, `feature`, `docs`, `chore`, `refactor`).
Add existing topic labels for every subject materially covered by the final title, body, and diff.
Ignore incidental references, checklist text, and verification boilerplate when choosing topics.
Remove topic labels only when the final scope no longer supports those labels.
Preserve size, priority, review, and automation labels. Do not create or rename labels unless the user requests that change.
Apply label additions and removals in a batch with the title/body update. Verify the resulting metadata.

Apply exactly one package label: `pkg:core`, `pkg:harness`, `pkg:clai2`, `pkg:evals`, `pkg:graph` or `pkg:clai`.
Use the catalog descriptions to choose. Tests and docs count toward the package they cover.
Use `pkg:core` for repository infrastructure. For changes spanning packages, choose the package whose users the change serves.

Labelling needs triage permission on the repo (Pydantic team members and their agents). If it
fails, quote the actual error rather than concluding you lack permission. Size labels are
applied automatically — don't set them.

## Mechanical text/data refreshes

A small mechanical text/data refresh changes no code, runtime behavior, CI logic, or substantive
instruction or review policy. A catalog-row refresh is one example.

For this case, skip local `pre-push-review` and any additional substantive review dispatch. Do not
ask for a second-opinion choice. Automatic configured CI and hosted reviews still run.

## Before you push
- Commit the exact state you intend to push. Leave nothing staged, unstaged or uncommitted unless
  the user's instructions override this.
- For all other diffs, run `pre-push-review` before each push. Address every finding, commit the
  fixes, and repeat the review until it returns no findings. This applies before the first PR push
  and between every later PR iteration.
- A `pre-push-review` verdict belongs to the diff it read. Any later commit voids it — re-run against
  the new diff instead of carrying the earlier pass forward, and name the commit range each verdict
  covers when you report it.
- Never force-push an open PR branch. Push follow-up commits so previous reviews remain valid;
  maintainers can squash them when merging.
- Attempt the push. If it fails, read the real error — do not preemptively decide you lack
  permission from a flag or setting.

## After you push — the loop
1. **Watch CI to a terminal state.** Don't idle. If it fails, diagnose: fix if the failure is
   yours; if it's a known flake or pre-existing on main, say so with evidence.
2. **Triage every comment** (bots and humans alike). For each one:
   - **Valid** → fix it, then reply saying what changed, and react 👍.
   - **Invalid** → reply explaining concretely why (with code evidence), and react 👎.
   - Never silently ignore a comment, and never resolve a thread without a reply.
3. **Escalate real trade-offs, don't guess.** If a comment needs a maintainer decision (a design
   choice, an API trade-off, a behavioral default), leave a comment containing: the background,
   your reasoning, the decision that needs making, the trade-offs (pros/cons of each option), and
   your recommendation. Then **poll every 30 minutes for a reply** and continue when it lands.
4. Repeat until CI is green and no comment is outstanding.

## When the loop completes — consider a deep `douwebot` review

The repo has two standards reviewers, and they are independent:

- **`CI Review`** runs automatically once the `CI` workflow succeeds on the PR's current head. It
  submits an `APPROVE` or `REQUEST_CHANGES` verdict and has the more rigorous process — severity
  scale, sub-agent fan-out, per-finding verification.
- **`douwebot`** runs only when the `douwebot` label is applied, on a stronger model. It posts
  inline findings and a formal `APPROVE` or `REQUEST_CHANGES` review, and it deletes the label when
  it finishes, so each application buys exactly one review of the diff as it stands at that moment.

Applying the label adds a second opinion; it does not suppress or replace `CI Review`.

- Skip additional-review selection for a mechanical text/data refresh defined above.

Once the loop above has terminated — CI green, every comment triaged — decide whether to apply it
before handing the PR back or requesting merge:

- **Apply it last, not early.** It won't re-run on later pushes, so a deep review of a
  still-moving PR is wasted money.
- **Use judgment on whether it's warranted.** Skip it when you're highly confident there's nothing
  left to catch (typo fixes, dependency bumps, mechanical chores). Apply it for substantive
  changes: new features, behavior changes, public API surface, non-trivial bug fixes — and
  user-facing docs, where it catches things like examples using outdated models. In between, weigh
  cost against risk; smaller PRs are cheaper to review, so lean toward applying when unsure.
- **How:** `gh pr edit <number> --add-label douwebot`. This requires triage permission on the repo
  (Pydantic team members and their agents). If it fails, quote the actual error — don't skip it
  based on an assumed lack of permission.
- **Known refusal:** the job fails without reviewing if the PR touches an `AGENTS.md` or `CLAUDE.md`
  at any depth, `CLAUDE.local.md`, `.mcp.json`, or anything under `.claude/`, `.agents/` or
  `agent_docs/` — a security guard against a PR editing the reviewer's own instructions. The guard
  is skipped for an author with write or admin access on the repo. Don't apply the label to a PR the
  guard covers; the red check is the guard working.
- **Afterwards, re-enter the loop.** The review posts inline comments and a formal verdict that
  need the same triage as any other review.

## Before handing the PR back

Run this final metadata check after CI, comments, and any selected `douwebot` review have settled:

1. Dispatch a fresh subagent that has not worked on the PR.
2. Give it the PR URL, linked issue, current `base...HEAD` diff, final test status, title, body, labels, and complete catalog.
3. Ask it to check only the title, body, and labels against this skill and the root `AGENTS.md`.
4. Require either `current` or exact corrections: replacement title/body and label additions/removals.
5. Apply every correction. Committed changes restart the post-push loop; GitHub metadata-only changes do not.
6. After a replacement, repeat the check with another fresh subagent.
7. Hand the PR back only after the check reports `current`.