chore(engineering-analytics): add rafapherding skill for driving PRs to green - #74105
Closed
rafaeelaudibert wants to merge 3 commits into
Closed
chore(engineering-analytics): add rafapherding skill for driving PRs to green#74105rafaeelaudibert wants to merge 3 commits into
rafaeelaudibert wants to merge 3 commits into
Conversation
…to green Captures the unattended PR-custody workflow: read the review bots, judge each finding accurate or not, reply to every one, fix what is real with a test that fails without the fix, and drive CI to green. Graphite stacks are worked bottom-up throughout, with a single submit at the end. Lives in engineering_analytics alongside the other CI skills. Two references carry the detail that does not belong in the entry point: the failure taxonomy that separates a real red check from a runner-starvation timeout or an aggregation gate, and what a reply to a bot should contain, with worked examples for accurate, out-of-scope, and wrong. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
rafaeelaudibert
added
the
skip-agent-review
…rding skill Both hit live while shepherding the stack the skill was written from, and neither was in it. A push cancels the run in flight for that branch, and the aggregation gates then report their CANCELLED dependencies as FAILURE. A wall of "Tests Pass" failures right after pushing means superseded, not broken, and a CI monitor that does not filter gates will wake you on every push. Adds the state-breakdown check that distinguishes them and the jq filter for real failures only. A stale test patch does not reliably fail. When the URL under test resolves for real, the test makes a live request and passes on whatever production returns, so the local suite going green proves nothing about whether every seam was found. Says to grep the repo for the old seam immediately after moving one, before running anything. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Unlike the hot-table and atomic=False policies, a BLOCKED verdict from the migration risk analyzer has no acknowledgement file: it has to be split. The two rules that catch people are that RunPython may not share a file with schema changes and RunSQL may not share a file with anything, since both hold their locks for the whole file. Records the one-shape-per-migration layout, the ordering constraints a split has to preserve, and the three commands that verify the result. Also notes that this surfaces late, because the job reporting it is one of the ones most often lost to runner starvation, so a migration can sit blocked across several pushes unseen. That is how it went here. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
Merging to
After your PR is submitted to the merge queue, this comment will be automatically updated with its status. If the PR fails, failure details will also be posted here |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
Getting a batch of open PRs to green is a repetitive job with a specific shape, and the parts that
go wrong are always the same parts. Review bots post a mix of real findings, true-but-out-of-scope
findings, and confident nonsense, and all three need a reasoned reply rather than a silent commit.
Red CI is mostly not your fault: runner-starvation timeouts and aggregation gates look identical to
real failures in
gh pr checks. And a stacked PR fixed in the middle gets clobbered by the nextrestack.
None of that was written down, so each pass rediscovers it.
Changes
Adds a
rafapherdingskill underproducts/engineering_analytics/skills/, next to the other CIskills.
The entry point covers the workflow: group the PR list into stacks, work each stack bottom-up
(
gt restack,gt up, onegt ssat the end), triage CI before reading a single comment, thenjudge each bot finding on mechanism, reachability, and scope before acting on it.
Two references carry the detail that would bloat the entry point:
reading-ci.md— the failure taxonomy. A job withsteps: 0and a duration exactly equal to itstimeout-minutesnever got a runner. A<Something> Tests Passjob failing in seconds isreporting a dependency. Also how to pull a real semgrep finding out of the log, and why a stamphog
REFUSEDis not a CI failure.replying-to-bots.md— what a reply should contain, with worked examples for accurate,out-of-scope, and wrong, plus the shape to use when a finding turns on a fact the repo cannot
answer.
Two things in there are load-bearing and worth calling out, because both cost me real time this
session:
Note
Every fix gets a test you watched fail. Revert the fix, run the test, confirm it fails, restore.
A security regression test that 401s before it reaches the code under test passes for the wrong
reason and proves nothing.
Warning
Changing how production code reaches the network moves the seam every test was patching. Tests
stop intercepting and start making real calls, which surfaces as a wall of DNS failures for
hostnames like
partner.example.com. Retarget the patch; don't revert the fix.How did you test this code?
hogli lint:skillspasses (125 skills, the 2 advisory warnings are pre-existing inturning-engineering-analytics-into-insightsand unrelated).hogli build:skillspicks uprafapherdingfromproducts/engineering_analytics/skills/rafapherding.No automated tests added. This is documentation with no runtime behavior, so there is nothing to
assert that would not just be re-reading the file. I did not test the skill by having a fresh agent
session follow it end to end.
Automatic notifications
Docs update
None needed, the skill is the doc.
🤖 Agent context
Autonomy: Human-driven (agent-assisted)
Claude Opus 5 wrote this in Claude Code while shepherding #73935, #74089 and #74095 in the same
session. The content is the workflow actually used on those three, not a guess at one, which is why
the references are specific about things like zero-step job timeouts and moved test seams.
Invoked
/writing-skillsfor the layout and naming rules, and/writing-code-commentsand/writing-testsearlier in the session while doing the PR work the skill describes.Two decisions worth flagging for review. First, placement: this is a dev-workflow skill rather than
a product-feature one, and
products/engineering_analytics/skills/already ownsinvestigating-ci-failuresanddiagnosing-ci-and-merge-bottlenecks, so it went there. Second, thename is a portmanteau I was asked for, which makes it opaque on its own, so the description carries
all the trigger terms and an explicit "not for" pointing at
diagnosing-ci-and-merge-bottlenecks. Worth a second opinion on whether an opaque name is a fairtrade against the shared skill-count budget, since every extra skill makes the others less likely to
fire.