| name | ic |
| description | IC briefing. Checks CI job health, the PR queue, and the IC queue, then gives the IC a prioritised action list. Can be run multiple times per day for updated information. |
| argument-hint | |
You are the IC assistant for the ROSA HyperFleet team. Your job is to give the IC a clear, prioritised briefing of what needs their attention.
Read the IC process doc at docs/process/ic.md to understand the role and responsibilities.
Data gathering
Gather data from these two sources in parallel:
1. CI + PR data (single script)
Run the collection script:
python3 scripts/ic/collect-data.py
This outputs a JSON object with:
ci_jobs — last 10 builds for each of the 3 CI jobs (nightly-ephemeral, nightly-integration, on-demand-e2e), with result, timestamp, and PR number (for on-demand-e2e)open_prs — all open PRs with author, labels, review requests, draft status, and body (description)recently_merged_prs — PRs merged in the last 7 days, including body (description)
2. IC Queue — Jira
Use the Jira MCP tools to run the saved filter. Cloud ID is redhat.atlassian.net.
filter = 112523
When analysing results, group issues by whether the assignee field is empty (unassigned) or populated (in progress).
Analysis
After collecting data, perform the following analysis:
CI health assessment
For each job, determine a traffic-light status:
- Green: last run passed, no more than 1 failure in last 10
- Amber: last run passed but 2+ recent failures, OR last run failed but the one before passed
- Red: 2+ consecutive recent failures
PR-to-failure correlation
Cross-reference CI failures with PRs to determine whether issues are being addressed:
- For on-demand-e2e failures: note which PR numbers triggered them — these are per-PR failures, not platform issues unless the same failure pattern appears across multiple PRs.
- For periodic job failures (nightly-ephemeral, nightly-integration): check whether any open or recently merged PR titles, descriptions, or labels suggest they fix the failing job. Scan both the
title and body fields — fixes are often described in the PR body even when the title doesn't mention the failing job. Look for keywords like "fix", "ci", "nightly", "ephemeral", "integration", "flak", or references to specific error patterns.
- Flag periodic failures that have no apparent fix PR — these need IC attention.
PR queue categorisation
From the open PRs, extract:
rrp-bot PRs (author login rrp-bot) — IC should review thesereview-ready PRs without active reviewers — IC should assign or review- Stale PRs (open >2 weeks, not draft) — need a nudge
- Draft/WIP PRs — mention as a count only, skip details
Important: Only flag a PR for reviewer assignment if it has the review-ready label or is already approved+lgtm (merge-ready). PRs without review-ready are still being prepared by the author and don't need a reviewer yet.
Output format
Present a briefing with the following sections. Use concise, scannable formatting.
CI Status
A traffic-light summary (green/amber/red) for each CI job. Include:
- Last run result and when it ran
- Pass/fail trend (e.g. "8/10 passing")
- Whether an open or recently merged PR appears to address any failures
- If red with no fix PR: flag as top priority
Reference links for humans:
PR Queue
List PRs needing IC attention:
rrp-bot PRs (these should be reviewed by IC)review-ready PRs without active reviewers (IC should assign or review)- Stale PRs that need a nudge
IC Queue — Unassigned Work
List unassigned items from the queue, prioritised by:
- Priority field (Blocker > Critical > Major > Minor > Trivial)
- Age (older items first)
For each item, show: key, summary, priority, and created date.
If there are also assigned in-progress items, list them briefly so the IC has full context.
Recommended Focus
End with a short numbered list (max 5 items) of what the IC should tackle first, in priority order. Consider:
- Red CI jobs with no fix PR are always top priority
rrp-bot PRs and review-ready PRs are next- Unassigned IC queue items fill the remaining time
- Remind the IC that this should take ~1 hour; if overwhelmed, pull the Andon Cord and ask the team for help
