Files
PaperClipAI/doc/PR-MERGE-REVIEW-WEBHOOK.md
T
DottaandPaperclip b6e9615aed feat(ci): request advisory merge review after PR verification (#19480)
## Thinking Path

> - Paperclip helps people manage AI agents and their work.
> - Pull requests already run CI and automated code review.
> - A completed check does not start a final advisory merge review.
> - A private routine webhook can connect these two steps.
> - The webhook credential must stay outside PR code and public logs.
> - This pull request adds an optional notification from trusted branch
code.
> - The routine records a recommendation. A person keeps merge control.

## Linked Issues or Issue Description

**What existing behavior does this improve?**

It connects successful PR verification to an advisory Paperclip routine.

**Subsystem affected**

GitHub Actions and the existing routine webhook API.

**Current behavior**

A PR can finish verification without asking a routine for a final merge
recommendation.

**Proposed behavior**

A trusted workflow reads the current PR and its complete check list. It
sends the PR identity to a private webhook when required checks pass and
other reported checks finish successfully, as skipped, or as neutral.
The receiving routine runs the configured merge review skill and records
its decision on its Paperclip task.

**Reason and benefit**

This gives maintainers an advisory decision after verification. It does
not grant merge control to the routine.

**Breaking changes**

None. Missing secrets or delivery errors leave verification results
unchanged.

Searched open and closed PRs and issues for related work. No duplicate
notification workflow was found. Related API work: #18058 and #18581.
This change uses the routine API and does not alter chat credential
setup.

## What Changed

- Add a default-branch workflow for PR workflow completions, external
checks, and commit statuses.
- Use read permissions and a GitHub environment restricted to master for
the private URL and bearer token.
- Require successful `ci / verify`, `ci / e2e`, and additional
branch-rule checks. Read every status page and refresh the PR before
delivery.
- Use a stable idempotency key for each PR, head, and base. Reject stale
events and changed candidates.
- Verify the two core check runs belong to the same real GitHub Actions
suite for the unchanged PR caller and master-owned reusable CI. Reject
forged names, renamed callers, and incomplete file listings.
- Keep HTTP redirects disabled, bound request time, and print fixed
delivery states only.
- Add regression checks and document private setup, pause, rotation, and
acceptance steps.
- Grant commit-status read access. Add a separate PR API probe with the
same read permissions and the real Actions token. The probe has no
private environment or webhook secrets and sends no notification.

## Verification

- Pass: `node --test
.github/scripts/tests/notify-pr-merge-review.test.mjs` (21 tests after
the security and token-permission fixes).
- Pass: actionlint v1.7.12 on `.github/workflows/pr-merge-review.yml`.
- Pass: live GitHub query of a current PR and its complete status list.
This found a GraphQL field mismatch during development. The corrected
query passed with 62 checks.
- Pass: the configured bearer webhook returned `test_received` with
`routineStarted=false` in setup mode. Readback confirmed an active
routine and environment secrets restricted to the master branch.
- Not run: `pnpm verify --explain` could not start because this checkout
has no installed workspace dependencies. Required CI will supply broad
coverage.
- The first quality-gate run rejected a CI-only title because this
change adds a script. The feature title passed the later gate.
- The security review found that PR-controlled check names could forge
eligibility. A focused regression failed before the fix and passed after
suite, run, caller, and reusable-workflow checks were added. The
corrected provenance query also passed against a completed historical CI
run without sending a webhook.
- Earlier CI at ec2c2dc failed in `workspace-recovery.test.ts` with a
generated company-prefix collision, in
`idle-domain-deadlines-postgres.test.ts` with a deadline snapshot
mismatch and timeout, and in `CoordinatorTasks.test.tsx` with a route
query mismatch. The browser shard also recorded a CSS preload test
failure before cancellation. These failures remain recorded in [run
42245040952](https://github.com/paperclipai/paperclip/actions/runs/42245040952).
This PR does not edit those production paths. Current master is merged
and fresh CI is required; the earlier failures are not claimed resolved.
- The repeated browser, general-server, and UI lanes passed at
715cee88ce. The earlier failures remain historical; their causes are not
all established. The Runner package lane was still running before the
next fix.
- A reviewer found missing `statuses: read`, which the broader local
token did not expose. The permission regression failed before the fix
and passed after it. The new read-only API probe will test the query
with the actual Actions token.
- Pass at c3c84394b7: the real Actions-token API probe, typecheck,
build, browser aggregate, security scan, and Greptile 5/5. No unresolved
review threads remain.
- The only remaining CI failure is a 15 second timeout in the existing
retained-native-session-cleanup-work backlog test (server shard 6/12).
This file and its production implementation are unchanged from the
previous candidate whose lane passed. The one scoped rerun passed at the
same head, and the required verify aggregate now passes. The timeout
remains in [run 42250393645, attempt
1](https://github.com/paperclipai/paperclip/actions/runs/42250393645/attempts/1);
its cause is not established. This is recovery evidence, not a claimed
root-cause fix.
- A nonblocking review suggestion to defer branch-rule lookup until a PR
is matched remains a minor efficiency follow-up. It is not a correctness
or security finding.
- Pending: a real notification and resulting routine task after this
workflow reaches master. The setup test did not start a PR review.

## Risks

- This workflow uses secrets. Keep the environment restricted to the
master branch. It runs only trusted branch code and does not read PR
artifacts, caches, or executable files.
- GitHub event timing and check visibility can race. Delivery is best
effort. A missed event can omit a review; it cannot change CI results or
merge a PR.
- Optional skipped and neutral checks permit an assessment. They do not
prove that a check ran. Required checks must succeed. The routine must
assess unavailable scanner coverage and other evidence gaps before its
recommendation.
- Add any new PR verification workflow name to the completion trigger
list.
- PRs that change the primary CI caller, `.github/workflows/pr.yml`,
need a manual advisory review until their caller change reaches master.
This prevents a PR from replacing verification with a forged result.
- This PR changes CI configuration. CODEOWNERS review and normal merge
rules still apply.

## Model Used

OpenAI GPT-6 family through Codex. The exact deployment model ID and
context window are not exposed to this run. Used reasoning, tool use,
and code execution.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with available version and
capability details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change and contains no internal
Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

---------

Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-11-06 15:25:57 -06:00

4.8 KiB

Advisory PR review notifications

The optional PR merge review notification workflow sends a private webhook after a current, open, non-draft PR targeting master passes verification. It does not approve or merge a PR. Webhook failures do not change PR CI results.

Private configuration

Create a GitHub Actions environment named pr-merge-review. Restrict its deployment branch policy to the master branch only, with no tag policy. Keep these two values in environment secrets, not repository variables, repository secrets, committed files, task comments, or workflow outputs:

  • PAPERCLIP_PR_REVIEW_WEBHOOK_URL: a Paperclip routine webhook URL.
  • PAPERCLIP_PR_REVIEW_WEBHOOK_TOKEN: its separate bearer signing secret.

Use a webhook trigger with signingMode: bearer. The routine should use always_enqueue concurrency so separate PRs cannot be coalesced into one task. Define required text variables pr_url, pr_number, head_sha, and base_sha. The notifier sends only those identifiers. The receiving agent must treat them as untrusted data and verify the repository, PR identity, and current revision.

Give the routine instructions to run pr-merge-review and publish its MERGE, HOLD, or DO NOT MERGE recommendation on the routine execution's Paperclip task. Require a fresh check of the head, base, reviews, required checks, and task gates. Keep GitHub approval, merge, auto-merge, and public-comment actions disabled in the instructions for this advisory phase. Configure the routine in Paperclip; do not put instance-specific URLs or credentials in this repository.

Delivery and verification

The notification workflow runs trusted default-branch code. It never checks out PR code or downloads PR artifacts or caches. It reacts to the repository's PR workflow completions, external check completions, and legacy commit statuses. When adding another PR verification workflow, add its name to workflow_run.

The separate PR review API access workflow checks notifier changes using the real read-only Actions token. It has no private environment or webhook credentials and sends no webhook. Both workflows explicitly grant actions, checks, contents, pull-requests, and statuses read access. The GraphQL status rollup needs both check and commit-status permissions.

The notifier reads the current PR's status rollup and branch rules. Both ci / verify and ci / e2e, plus any additional required checks, must succeed. Other checks must succeed or finish as skipped or neutral. Optional skipped or neutral checks permit an assessment; they do not prove successful verification. For example, a neutral scanner can mean coverage is unavailable. The advisory routine evaluates these evidence gaps and reviews before recommending merge. Workflow names do not exclude checks. The notifier and CI profile exporter run on the default branch, so their own checks do not belong to the PR revision.

Check names alone are not trusted. Both core aggregates must belong to the same GitHub Actions suite and a successful run of the repository's actual .github/workflows/pr.yml. Its run metadata must reference pr-trusted.yml@master. The complete PR file list must prove that the caller pr.yml was not changed or renamed. A missing or truncated list stops delivery. PRs that change this caller need a manual advisory review until their caller change reaches the trusted branch. Other workflow changes can still be reviewed.

A final refresh rejects head or base changes during evaluation. Paperclip's idempotency receipt deduplicates events for the same PR, head, and base. A new base can produce a new review. The routine must refresh again before its final recommendation. Event delivery and host check visibility can race; notifications are best effort, with no polling or delivery guarantee.

The HTTP request uses TLS, a bearer header, no redirects, and a 15 second timeout. Logs contain fixed delivery states only. Request and response bodies, URLs, tokens, and raw exceptions are never logged. Missing secrets skip delivery.

Run the focused regression checks with:

node --test .github/scripts/tests/notify-pr-merge-review.test.mjs

For setup verification, use Paperclip's webhook setupPending test mode before activation. This verifies authentication and ingress without starting a review. After the workflow is merged to master, inspect a real successful workflow delivery and its linked routine task to prove end-to-end operation. A saved configuration or a setup test alone does not prove that journey.

To pause reviews, pause the routine in Paperclip or disable its trigger. To stop notifications, disable this workflow or remove the two environment secrets. Rotate the trigger signing secret in Paperclip and replace the token environment secret together. Keep the URL secret even though it has no embedded bearer token.