Github:last Week

Last Week Report

Deep weekly analysis of repository health: issue investigation, PR status, CI trends.

When to Use

• Weekly standup preparation
• Sprint retrospective input
• Tracking project health over time

Auto-approved: All gh read commands are auto-approved.

Workflow

Phase 0: Gather Data (MUST run first)

Run the data gathering script to get a consistent snapshot of all issues, PRs, and CI runs. This prevents stale data from long-running analysis.

./.github/scripts/reports/weekly-report-data.sh 7 kagenti/kagenti

This creates JSON files in /tmp/kagenti/github/data/:

• merged-prs.json — PRs merged in the period
• open-prs.json — Currently open PRs (live snapshot)
• open-issues.json — Currently open issues (live snapshot)
• new-issues.json — Issues created in the period
• ci-runs.json — CI runs on main
• ci-runs-all.json — All CI runs

IMPORTANT: All subsequent phases MUST read from these JSON files, NOT re-query gh. This ensures consistency — PRs won't change state mid-analysis.

Phase 1: Quick Stats

Read from the gathered data:

echo "Merged PRs: $(jq length /tmp/kagenti/github/data/merged-prs.json)"

echo "Open PRs: $(jq length /tmp/kagenti/github/data/open-prs.json)"

echo "Open issues: $(jq length /tmp/kagenti/github/data/open-issues.json)"

echo "New issues: $(jq length /tmp/kagenti/github/data/new-issues.json)"

Phase 2: Deep Issue Analysis

For EVERY open issue (from open-issues.json), investigate:

Read from gathered data (do NOT re-query gh):

cat /tmp/kagenti/github/data/open-issues.json | jq '.[].number'

For each issue, determine:

1. Still relevant? — Check if the bug still exists on upstream/main

   • Search codebase for the affected code/component
   • Check if a fix was merged since the issue was created
   • Check if a test covers this scenario

2. Work in progress? — Check if any PR references this issue

   gh pr list --repo kagenti/kagenti --state all --search "issue-number" --json number,title,state
   

3. Severity classification:

   • Security: Credential leaks, auth bypass, injection
   • Blocking: Breaks deployment, tests, or user workflow
   • Bug: Broken behavior but workaround exists
   • Feature: New capability request
   • Epic: Multi-PR initiative (track progress)
   • Stale: No activity >30 days, may be outdated

4. Actionable recommendation: Close, update, assign, or create PR

Phase 3: Deep PR Analysis

For EVERY open PR (from open-prs.json), investigate:

cat /tmp/kagenti/github/data/open-prs.json | jq '.[].number'

For each PR, determine:

1. CI status: Does it pass all checks?
2. Needs /run-e2e?: Does it touch kagenti/charts/deployments/.github paths?
3. Review status: Approved, changes requested, or waiting?
4. Health: Is it stale (>14 days)? Has merge conflicts?
5. Summary: What does this PR do? (read the body)

Classify each PR:

Phase 3b: CI Failure Timeline

Get all runs on main (last 7 days):

gh run list --repo kagenti/kagenti --branch main --limit 30 --json databaseId,conclusion,name,createdAt,headSha,displayTitle

For each failed run, get the failing job and step:

gh run view <run-id> --repo kagenti/kagenti --json jobs --jq '.jobs[] | select(.conclusion == "failure") | "\(.name): \(.steps[] | select(.conclusion == "failure") | .name)"'

Get the commit that triggered the failure:

gh run view <run-id> --repo kagenti/kagenti --json headSha --jq '.headSha'

Check what that commit changed:

gh api repos/kagenti/kagenti/commits/<sha> --jq '.files[].filename'

Build a timeline of failures with root cause correlation:

• Map each failure to the triggering commit
• Identify if the same job/step fails repeatedly (infrastructure issue)
• Identify if failures started after a specific merge (regression)

Since E2E doesn't run on every merge, find candidate PRs that could have caused a failure:

gh pr list --repo kagenti/kagenti --state merged --json number,title,mergedAt,changedFiles --jq '.[] | select(.mergedAt > "LAST_SUCCESS_DATE" and .mergedAt < "FAILURE_DATE") | "#\(.number) \(.title) (\(.changedFiles) files)"'

For each candidate PR, check if it touched relevant paths:

• charts/ or deployments/ → likely deploy/config issue
• kagenti/tests/ → test change may have introduced failure
• .github/workflows/ or .github/scripts/ → CI infrastructure change
• kagenti/backend/ or kagenti/auth/ → application logic change

Correlate the failure type with the PR changes to identify the most likely cause.

Phase 4: Generate Report

Write the full report as markdown. Structure:

# Weekly Report: [date range]

## Quick Stats
- Merged PRs: N
- New issues: N
- CI pass rate: X/Y on main

## Issue Analysis

### Security Issues
| # | Title | Status | Recommendation |
...

### Blocking Issues
| # | Title | Status | Recommendation |
...

### Bug Reports
| # | Title | Still affects main? | PR exists? | Recommendation |
...

### Feature Requests
| # | Title | Priority | Recommendation |
...

### Epics (track progress)
| # | Title | PRs merged/open | % done estimate |
...

### Issues to Close
| # | Title | Reason |
...
(Issues where fix was already merged, or issue is outdated)

## PR Analysis

### Ready to Merge
| # | Title | Author | Approved by |
...

### Needs Review (CI passing)
| # | Title | Author | Days waiting |
...

### Needs /run-e2e
| # | Title | Reason |
...

### Unhealthy PRs
| # | Title | Problem | Next step |
...

## CI Health

### Pass Rate
- Main branch: X/Y runs passed (Z%)
- HyperShift E2E: X/Y passed
- Kind E2E: X/Y passed
- Security/Lint/Build: X/Y passed

### Failure Timeline
| Date | Workflow | Job | Failure | Likely Cause |
|------|----------|-----|---------|-------------|
| ... | E2E HyperShift | Deploy & Test | timeout | Cluster creation slow |
| ... | CI | build | compile error | PR #N broke import |

### Failure Patterns
- **Recurring**: [failures that happen repeatedly — infrastructure, flaky tests]
- **One-off**: [failures tied to specific commits]
- **Trend**: [getting better/worse over time]

### Root Cause Correlation
For each failure, identify candidate PRs merged between last success
and the failure. Check which files they touched to find likely cause.

## Action Items

Single flat list, ordered by priority. No timescales.

| # | Action | Owner | Priority |
|---|--------|-------|----------|
| 1 | [Highest priority action] | [owner] | P0 |
| 2 | [Next action] | [owner] | P1 |

Save report:

# Save to /tmp/kagenti/github/weekly-report.md

Phase 5: Ask User

After generating the report, ask:

Weekly report ready. Want me to create a GitHub issue with this? Suggested title: "📊 Weekly Report [start-date] – [end-date]"

You can also suggest a different title.

Only create the issue after user confirms title and content.

Related Skills

• github:issues - Deep dive into individual issue triage
• github:prs - Deep dive into individual PR health
• git:status - Worktree and PR status overview
• ci:status - Detailed CI check analysis
• rca:ci - Investigate CI failures
• repo:issue - Issue creation format