Create Tekton Task/Pipeline for /retest slash command

[vdemeester]

Parent

Part of #3121 — Migrate slash commands from GitHub Actions to dogfooding cluster

Summary

Create a Tekton Pipeline that implements the /retest command, replacing the GitHub Actions workflow _chatops_retest.yml.

Current Behavior (GHA)

The current workflow:

1. Gets the PR's HEAD commit SHA
2. Lists all check runs for that commit
3. For each unique run ID, reruns failed jobs using gh run rerun <id> --failed
4. Comments on failure, adds 🎉 reaction on success

Proposed Tekton Implementation

Pipeline Structure

Pipeline: slash-retest
├── Task: retest
│   ├── Get PR HEAD commit SHA via gh API
│   ├── List check runs for the commit
│   ├── Filter to failed runs
│   ├── Rerun each failed run via `gh run rerun`
│   └── Add reaction to trigger comment
└── Finally Task: report-status (on failure)
    └── Comment on PR with error details

Key Considerations

• Simplest of the three: Just API calls, no git operations needed
• GitHub API only: Uses gh CLI to interact with GHA — no repo checkout needed
• Permission model: The current GHA also allows the PR author to /retest (not just write-access users). The slash routing enforces "write" but the retest workflow itself is more permissive. We may want to keep this behavior.
• No workspace needed: Just needs a container with gh CLI and the CHATOPS_TOKEN

Webhook Payload Fields Needed

From issue_comment event:

• body.repository.full_name — repo to rerun checks for
• body.issue.number — PR number (to get HEAD SHA)
• body.issue.pull_request.url — PR API URL
• body.comment.id — for adding reaction

Files to Create

• tekton/ci/slash-commands/retest-pipeline.yaml
• tekton/ci/slash-commands/retest-trigger.yaml

Acceptance Criteria

• /retest reruns all failed GHA check runs for the PR's HEAD commit
• Adds 🎉 reaction on success
• Comments on the PR if something goes wrong
• Handles PRs with no failed runs gracefully

[sahilleth]

This looks really interesting, especially the migration of slash command handling into Tekton-based workflows.

I’ve been working on a related PR in pipeline around surfacing slash commands, and I’d like to contribute here as well.

Would it be helpful if I start with implementing the /retest Tekton Task/Pipeline (#3126) as a first step?