Skip to content

ci(fork): mirror only, without the paused gate or review requests - #91

Merged
jadoonf merged 1 commit into
mainfrom
garnet/mirror-only-twin
Oct 7, 2026
Merged

jadoonf merged 1 commit into
mainfrom
garnet/mirror-only-twin

Conversation

@jadoonf

@jadoonf jadoonf commented Oct 7, 2026 •

Copy link
Copy Markdown

Summary

Fork-only CI change. The evidence mirror keeps copying the head-bound Garnet comment into the PR description from the default branch (no PR code runs). It no longer publishes garnet/evidence and no longer posts reviewer trigger comments or Devin API requests. Reviewers see the mirrored record on their normal first review, e.g. when a draft that recorded during CI is marked ready.

  • garnet-evidence-mirror.yml: drop the gate and re-review steps, drop checks: write and the reviewer/Devin env, and also follow the Garnet Bootstrap Routing recorder.
  • garnet-evidence-gate.yml: paused (manual workflow_dispatch only). Live records do not declare capture completeness yet, so it cannot pass.
  • garnet-evidence-mirror.mjs: the mirrored preamble no longer prescribes a citation line. It keeps the facts-only / judgment-stays-with-the-reviewer note.

garnet-rereview.mjs and garnet-evidence-gate.mjs stay in the tree, so this is easy to revert.

Squash Commit Body

The gate cannot pass until records declare capture completeness, so it runs by
hand only. Reviewers read the mirrored record on their normal first review;
nothing posts trigger comments. The mirror also follows the bootstrap routing
recorder.

Checklist

  • Fork-only workflows; no published package changes.
  • actionlint clean on both workflows.

Link to Devin session: https://app.devin.ai/sessions/b554fb265c3347cfb2b92892896f618d
Open in Devin Desktop: https://app.devin.ai/desktop/session/b554fb265c3347cfb2b92892896f618d?variant=devin
Requested by: @jadoonf


View with [code]smith Autofix with [code]smith
Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.


Note

Low Risk
Fork-only GitHub Actions wiring; no runtime package or auth changes, though branch protection that required garnet/evidence will no longer be satisfied automatically until the gate is re-enabled.

Overview
Fork CI narrows Garnet evidence automation to description mirroring only—no garnet/evidence check and no bot/Devin review triggers from the mirror workflow.

The evidence mirror workflow now listens to Garnet Bootstrap Routing completions as well as TS CI, drops checks: write and reviewer/Devin env, and removes the steps that ran garnet-evidence-gate.mjs and garnet-rereview.mjs. Reviewers are expected to see the mirrored record on a normal first review (e.g. draft → ready).

The garnet/evidence gate workflow is paused for automatic workflow_run / issue_comment runs until records declare capture completeness; it only runs via workflow_dispatch with a required PR number (manual path still supports the old event shapes in the job if).

The mirror script preamble no longer includes a prescribed citation line or the blockquoted runtime-evidence format—only the facts-only / judgment-stays-with-reviewer note remains.

Reviewed by Cursor Bugbot for commit 76ad85d. Bugbot is set up for automated code reviews on this repo. Configure here.

Runtime evidence (Garnet)

No runtime evidence is bound to head 76ad85d yet. The sticky Garnet Runtime Review comment either has not been posted for this head or describes an earlier commit. Missing evidence means no record, not a clean run.

The gate cannot pass until records declare capture completeness, so it runs by hand only.
Reviewers read the mirrored record on their normal first review; nothing posts trigger
comments. The mirror also follows the bootstrap routing recorder.
@devin-ai-integration

Copy link
Copy Markdown

I'll fix CI failures and address comments from users with write access that start with 'Devin'.

  • Disable automatic comment, CI, and merge conflict monitoring

@coderabbitai

coderabbitai Bot commented Oct 7, 2026 •

Copy link
Copy Markdown

Review in Change Stack →

Note

Currently processing new changes in this PR. This may take a few minutes, please wait...

⚙️ Run configuration
  • Configuration used: Repository: garnet-labs/pnpm/.coderabbit.yaml
  • Review profile: CHILL
  • Plan: Advanced
  • Run ID: eb251e13-51af-4c4c-8242-77bc4d6bc565
📥 Commits

Reviewing files that changed from the base of the PR and between fc1155e and 76ad85d.

📒 Files selected for processing (3)
  • .github/scripts/garnet-evidence-mirror.mjs
  • .github/workflows/garnet-evidence-gate.yml
  • .github/workflows/garnet-evidence-mirror.yml
 ______________________________________
< I am an equal opportunity nitpicker. >
 --------------------------------------
  \
   \   \
        \ /\
        ( )
      .( o ).
✨ Finishing Touches
📝 Generate docstrings
  • Commit to this branch
  • Create a new PR
🧪 Generate unit tests (beta)
  • Commit to this branch
  • Create a new PR
  • Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

@jadoonf
jadoonf marked this pull request as ready for review October 7, 2026 20:17
@jadoonf
jadoonf merged commit 7243d0e into main Oct 7, 2026
20 of 23 checks passed
@qodo-code-review

Copy link
Copy Markdown

PR Summary by Qodo

Make Garnet evidence mirroring independent of the paused gate

⚙️ Configuration changes 🕐 10-20 Minutes

Grey Divider

AI Description

• Mirror head-bound evidence after TS CI or Garnet Bootstrap Routing without publishing a check or
 requesting reviews.
• Run the evidence gate only by manual dispatch until records declare capture completeness.
• Remove prescribed citation wording while keeping the reminder that evidence does not replace
 reviewer judgment.
Diagram

graph TD
  Recorders["Recorder workflows"] --> Mirror["Evidence mirror"] --> Description["PR description"]
  App["Garnet App comment"] --> Mirror
  Dispatch["Manual dispatch"] --> Gate["Evidence gate"] --> Check["Head check"]
Loading
High-Level Assessment

The following are alternative approaches to this PR:

1. Keep the gate automatic but nonrequired
  • ➕ Preserves continuous visibility of gate results.
  • ➖ Produces failing checks while live records lack capture completeness.
  • ➖ Does not meet the goal of leaving the mirror check-free.

Recommendation: Keep the PR's split: automatic mirroring gives reviewers the record without generating failing checks or unsolicited review requests, while manual dispatch preserves the gate for verification when needed. An automatic informational gate is less useful until records can satisfy its completeness contract.

Files changed (3) +18 / -45

Documentation (1) +1 / -3
garnet-evidence-mirror.mjsRemove the prescribed citation line from mirrored evidence +1/-3

Remove the prescribed citation line from mirrored evidence

• The generated preamble no longer tells reviewers to use a specific citation format. It retains the facts-only framing and leaves judgment with the reviewer.

.github/scripts/garnet-evidence-mirror.mjs

Other (2) +17 / -42
garnet-evidence-gate.ymlPause automatic evidence checks +11/-8

Pause automatic evidence checks

• Replaces workflow and comment triggers with a manual dispatch that accepts a PR number. The gate remains available, but no longer runs automatically while live records lack capture completeness.

.github/workflows/garnet-evidence-gate.yml

garnet-evidence-mirror.ymlLimit automatic CI to mirroring evidence +6/-34

Limit automatic CI to mirroring evidence

• Adds Garnet Bootstrap Routing as a recorder trigger and removes the check-publication and review-request steps. Drops the associated check permission, reviewer settings, and Devin token.

.github/workflows/garnet-evidence-mirror.yml

@devin-ai-integration devin-ai-integration Bot left a comment •

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

👀 1 finding needs your review

Devin fixed 1 of 2 findings on 76ad85d. Click a finding below to jump to its comment.

For your review (1)

Fixed by Devin (1)

  • Gate header describes a paused trigger as active

View all findings in Devin Review

Devin Review

@qodo-code-review

Copy link
Copy Markdown

Code Review by Qodo

🐞 Bugs (1) 📘 Rule violations (0) 📜 Skill insights (0)

Grey Divider


Remediation recommended

1. Withdrawn evidence can leave a green check 🐞 Bug ☼ Reliability
Description
Removing the issue_comment trigger means the gate no longer re-evaluates an existing
garnet/evidence check when the Garnet App edits its record. If a check has succeeded on a PR head
and that record is later changed to pending or incomplete, the successful check remains on that head
until someone manually dispatches the gate.
Code

.github/workflows/garnet-evidence-gate.yml[R14-19]

+on:
+  workflow_dispatch:
+    inputs:
+      pr_number:
+        description: Pull request number
+        required: true
Relevance

●● Moderate

Intentional manual pause removes active gating, but stale historical checks remain a plausible
reliability concern.

PR-#91
PR-#85

ⓘ Recommendations generated based on similar findings in past PRs

Evidence
The workflow now has only a manual trigger. The gate script detects an incomplete or missing bound
record when invoked, but publishes a replacement check run only during that invocation; it does not
expire a prior successful run on its own. This is a static inference about the case where a
successful check already exists on the head.

.github/workflows/garnet-evidence-gate.yml[12-19]
.github/scripts/garnet-evidence-gate.mjs[75-145]
.github/scripts/garnet-evidence-gate.mjs[219-245]

Agent prompt
The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
A successful `garnet/evidence` check can remain green after its supporting record is withdrawn because the gate no longer runs on comment edits.

## Fix Focus Areas
- .github/workflows/garnet-evidence-gate.yml[12-19]
- .github/scripts/garnet-evidence-gate.mjs[219-245]

## Recommended Fix
Keep the gate paused for new automatic publications, but re-evaluate an already-published check when the trusted Garnet comment changes, publishing a non-success result when its record is no longer valid. Alternatively, retire existing checks from any required-check configuration while automatic revalidation is paused.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

Dismiss ↗ | View ↗


Grey Divider

Context sources
✅ Compliance rules (platform): 36 rules
✅ REVIEW.md
Review mode: Auto: ⚖️ Balanced: CI workflow and permission changes carry security and behavioral risk.

Grey Divider

Tip of the day
💡 Did you know, you can route each severity your way: inline, summary, both, or drop

More tips ↗ | Customize Qodo ↗ | Qodo docs ↗

Grey Divider

Qodo Logo

@qodo-code-review

Copy link
Copy Markdown

PR Code Suggestions ✨

Warning

/improve is deprecated. Use /agentic_review instead (removal date not yet scheduled).

No code suggestions found for the PR.

@garnet-runtime-review

Copy link
Copy Markdown

Execution Profiles for 76ad85d · recording

⏳ updates in place as each job's record lands


💡 How to read this
Runner.Worker          ← process on a path
└─ npm
   └─ node             ← process that acted
      └─ ○ npmjs[.]org ← observed action

follow a path downward to see what ran and what it did — each path to an observed action is an execution chain

names on the path = processes · ○ = observed action · (…) = context

@cursor cursor Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

Fix All in Cursor

❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.

Want higher recall? High effort reviews run extra passes and find more bugs. A team admin can switch effort levels in the Cursor dashboard.

Reviewed by Cursor Bugbot for commit 76ad85d. Configure here.

on: # zizmor: ignore[dangerous-triggers] default branch only, no pull request code runs here
workflow_run:
workflows: ["TS CI"]
workflows: ["TS CI", "Garnet Bootstrap Routing"]

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Mirror trigger names missing workflow

Medium Severity

The workflow_run trigger lists Garnet Bootstrap Routing, but no workflow in this repository uses that name. GitHub never emits workflow_run for an unknown workflow, so the mirror does not follow that recorder after it completes.

Fix in Cursor Fix in Web

Reviewed by Cursor Bugbot for commit 76ad85d. Configure here.

@devin-ai-integration devin-ai-integration Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Devin Review found 1 potential issue.

Devin Review

on: # zizmor: ignore[dangerous-triggers] default branch only, no pull request code runs here
workflow_run:
workflows: ["TS CI"]
workflows: ["TS CI", "Garnet Bootstrap Routing"]

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟡 Bootstrap completions never refresh evidence

When bootstrap routing finishes, Garnet Bootstrap Routing cannot trigger the mirror because no workflow has that name. Without a matching comment event, the PR description keeps its previous evidence.

Learn more

A workflow_run trigger matches the name of an existing workflow in this repository. No checked-in workflow is named Garnet Bootstrap Routing, so that completion cannot start the mirror. The separate comment trigger can still refresh the description when a trusted Runtime Review comment is created or edited, but it does not make the new completion trigger functional.

Example: A bootstrap-only run finishes while its Runtime Review comment has not generated an eligible comment event. The mirror does not start for that completion, and the description retains the prior evidence rather than refreshing from the completed run.

Recommended fix: Identify the actual workflow name of the bootstrap recorder on the default branch and use it in the workflows list. If the recorder has no workflow in this repository, remove this inert trigger or add the intended recorder before relying on completion events.

Devin Review


Was this helpful? React with 👍 or 👎 to provide feedback.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant