Auto Review Bot intermittently fails with "no artifacts found" (recurrence of #1008, now repo-wide)
- Dominant language
- Solidity
- Stars
- 756
- Forks
- 1.1k
- Avg merge
- 2d 20h
- Merged PRs (30d)
- 14
Description
The Auto Review Bot workflow has been failing intermittently for several days, always at the same step with the same error, across many unrelated PRs — not isolated to one PR or one contributor.
**Error:** `Fetch PR Number` step fails with `no artifacts found` when `dawidd6/action-download-artifact` tries to pull the `pr-number` artifact from the corresponding `auto-review-trigger.yml` run, even though that trigger run itself reports `success`.
**Confirmed recent, still-recurring instances** (checked directly via `gh run list`/`gh run view`, not just the notification):
- Run 34004830108 (2026-09-06T01:48Z) — trigger: "Add ERC: Signed Service Payment Quotes" (#1990)
- Run 33965532752 (2026-09-05T12:15Z) — trigger: "Update ERC-7730: define wallet behavior for non-zero native value"
- Run 33953718188 (2026-09-05T07:52Z) — trigger: same head
- Run 33932016327 / 33931967452 (2026-09-05T00:08-00:09Z) — triggers: an email-titled comment run and "Add ERC: Portable Execution Plan Artifact"
- Run 33891640763 / 33891597899 (2026-09-04T15:48-15:49Z) — different head entirely
30+ failed runs since 2026-09-03 by `gh run list --workflow "Auto Review Bot" --status failure`, interleaved with genuine successes on the exact same commit SHA at other times — so it's intermittent/racy, not a hard break in the workflow definition itself.
**This is a recurrence, not new territory:** #1008 (2025-10-14) hit the identical `no artifacts found` error on PR #930 and was closed only because that PR got merged anyway, not because the underlying cause was found or fixed. Nearly a year later it's back and repo-wide, so whatever closed #1008 didn't address the actual race/timing condition between the trigger workflow's artifact upload and the downstream `workflow_run` listener picking it up.
Net effect: the Auto Review Bot silently doesn't run on an unpredictable subset of PR comments/updates, with no visible signal to the PR author or reviewer that a review was expected but never happened.
Contributor guide
Research direction
Start by inspecting the auto-review-trigger.yml and Auto Review Bot workflow definitions, then use `gh run list` and `gh run view` on the recent failing and successful runs. Trace when the `pr-number` artifact is uploaded and fetched by the `workflow_run` listener. Done means the bot reliably obtains the artifact and runs reviews for affected PR comments and updates.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- github-actions
- Domain
- ci-cd, devops
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 48/100