Pilot Update

Quay Agentic SDLC Pilot Update

Gather data from Jira and GitHub, draft a biweekly update comment, get user input on subjective sections, and post the final comment to PROJQUAY-11352.

Tracking issue: PROJQUAY-11352 ("Quay Agentic Tool Pilot") Deadline: June 30, 2026

Prerequisite: This skill uses Jira MCP tools (mcp__mcp-atlassian__jira_*) which are configured at the Ambient session level, not in the repo's .mcp.json. It must be run inside an Ambient Code session with the Jira integration enabled. If running locally, the skill falls back to the Jira REST API via curl using the same credential locations as .claude/scripts/jira-ops.sh (~/.config/acli/jira_config.yaml + token from ~/.config/acli/token.txt or ~/.acli-token, or JIRA_API_TOKEN environment variable).

Phase 1: Gather Jira Activity (BFS traversal)

Determine the lookback window. Default is 14 days. If $ARGUMENTS was provided and is a single positive integer (> 0), use that value instead. If $ARGUMENTS is empty, missing, or not provided, use 14. If it contains a non-numeric value, zero, a negative number, or multiple tokens, inform the user the input is invalid and ask them to provide a single positive integer before proceeding.

Compute:

• end_date = today's date
• start_date = today minus lookback days

BFS traversal starting at PROJQUAY-11352:

Use a queue-based BFS. Start with ["PROJQUAY-11352"]. Before traversal, fetch and store PROJQUAY-11352's own fields using mcp__mcp-atlassian__jira_get_issue with fields="key,summary,description,status,assignee,issuetype,created,updated".

For each key in the queue:

1. Search for direct children: parent = KEY ORDER BY updated DESC
   • Use mcp__mcp-atlassian__jira_search with fields="key,summary,description,status,assignee,issuetype,created,updated" and limit=50
   • Paginate with start_at if total > 50 (pagination applies per parent-key search, not once globally)
2. Fetch each child's key, summary, description, status, assignee, issuetype, created, updated fields
3. Add each child's key to the queue
4. Continue until the queue is empty

The hierarchy under PROJQUAY-11352 is at least 3 levels deep (Feature → Epic → Story/Task). The BFS must traverse all levels.

After collecting all keys in the hierarchy (including PROJQUAY-11352 itself, with fields stored for each):

Fetch changelogs for all discovered keys using mcp__mcp-atlassian__jira_batch_get_changelogs with fields="status" to get status transitions. Each changelog item has a created timestamp — compare it (in UTC) against the lookback window using inclusive start, exclusive end: [start_date 00:00 UTC, end_date+1 00:00 UTC). If more than 20 keys, batch in groups of 20.

Classify findings:

• Completed issues: status transitioned to "Done", "Closed", "Resolved", or "MODIFIED" within the window
• Created issues: created field is within the window
• Status transitions: any status change within the window (excluding completions)
• Blocked issues: status is "Blocked" or summary/description contains "blocked" (case-insensitive)

Phase 2: Gather GitHub Activity

Repos to search:

• quay/quay
• quay/quay-operator
• quay/quay-bridge-operator
• quay/quay-bridge-operator-konflux
• quay/container-security-operator
• quay/quay-builder
• quay/quay-builder-qemu
• quay/quay-distribution
• quay/quay-konflux-components
• quay/quay-fbcs
• quay/registry-proxy
• quay/registry-proxy-tests
• quay/mirror-registry
• quay/mirror-registry-konflux
• quay/quay-tests

All repos are under the quay/ org and team-owned. No cross-team author filtering is required.

For each repo, search for PRs using the GitHub CLI:

Merged PRs (last N days):

gh search prs --repo quay/<repo> --merged --merged-at ">={start_date}" --json number,title,author,closedAt,url --limit 50

Open PRs (in progress):

gh search prs --repo quay/<repo> --state open --updated ">={start_date}" --json number,title,author,createdAt,url --limit 50

Filter to pilot-relevant PRs by checking if any of these apply:

• Has label agentic-sdlc-pilot or ai-pilot
• Touches paths: .claude/, CLAUDE.md, AGENTS.md, WORKFLOW.md, claude-plugin/, skills, agents
• Title or body mentions: "SDLC", "pilot", "agentic", "skill", "claude", "AI", "ambient", "factory"
• PR body contains AI-generated signatures: Co-Authored-By: Claude, Generated with Claude Code, Generated with Ambient

If a repo returns a 404 or 403, skip it gracefully and note it was skipped in the output.

Collect for each pilot-relevant PR: number, title, state, merged_at or created_at, repo name, author.

Phase 3: Draft the Update

Assemble gathered data into the 5-section template below.

Section 1: What We Tried (auto-populated)

Populate from:

• New Jira issues created in the window (list as: [KEY] Summary — issuetype)
• PRs opened (not yet merged) — include repo and PR number
• Tools, workflows, or skills referenced in PR titles/descriptions/Jira summaries
• Explicitly note which tool was used where possible: Ambient Code, Factory AI, or manual
• Note the workflow type: bug fix, story implementation, triage, documentation, CI/CD
• Note the issue type: Bug, Story, Epic, Task

Section 2: What Happened (auto-populated)

Populate from:

• PRs merged during the window — include repo, PR number, title, and link
• PRs opened but not yet merged — note as "in review" or "in progress"
• PRs closed without merging — note as "closed/abandoned" with reason if available
• Jira issues completed (status → Done/Closed/Resolved/MODIFIED) within the window
• Status transitions on pilot-related issues
• CI/test results if notable (failures, regressions)

Do NOT list PRs that are ordinary feature or bug work unless they were produced using the AI-assisted workflow. The signal here is: did the agentic pipeline produce output, and did that output land?

Section 3: What We Learned (seeded + user input)

Seed with data where possible:

• PR review comment themes (e.g., recurring feedback patterns on AI-generated PRs)
• Issues that required significant rework or were rejected
• Issues that went smoothly and why (fast merge, no review comments)

After presenting the seeded bullets, use AskUserQuestion to prompt the user:

The "What We Learned" section above is seeded from PR/Jira data. Please add subjective insights: What surprised you? What context was missing? What would you do differently? Any observations about code quality, review burden, or team dynamics?

Section 4: What's Blocked (seeded + user input)

Seed with data:

• Jira issues with "Blocked" status
• Issues whose descriptions mention: blocked, waiting, dependency, access, permission
• PRs stuck in review for >7 days

After presenting the seeded bullets, use AskUserQuestion to prompt the user:

The "What's Blocked" section above is seeded from Jira/GitHub data. Please add any additional blockers: tooling gaps, access issues, Jira integration problems, anything else slowing you down?

Section 5: What We're Trying Next (seeded + user input)

Seed with data:

• Jira issues in "New", "To Do", or "Refinement" status under PROJQUAY-11352
• Unmerged pilot-relevant PRs (work in flight)
• Upcoming epics or stories with future start dates

After presenting the seeded bullets, use AskUserQuestion to prompt the user:

The "What We're Trying Next" section above is seeded from upcoming Jira work. Please add or adjust: planned experiments, focus areas, changes in strategy or approach?

Phase 4: Review and Confirm

Assemble the complete comment using the template below for terminal preview (Markdown for readability). The actual posted comment will use ADF format — see Phase 5.

## Quay Agentic SDLC Pilot Update — {start_date} to {end_date}

### What We Tried
{bullet list: tools used, workflows attempted, issue types worked on, PRs opened}

### What Happened
{bullet list: PRs merged/closed, issues completed, status changes, outcomes}

### What We Learned
{bullet list: surprises, missing context, quality observations, process insights}

### What's Blocked
{bullet list: tooling gaps, access issues, integration problems, dependencies}

### What We're Trying Next
{bullet list: upcoming work, planned experiments, strategy changes}

Display the full formatted Markdown in the terminal. Use AskUserQuestion to ask the user to confirm or request edits:

Here is the full draft comment for PROJQUAY-11352. Reply with:

• "post" to post as-is
• Specific edits to apply
• "cancel" to abort

Apply any requested edits and re-display until the user confirms with "post".

Deduplication rule: Each PR or Jira issue should appear in at most one of "What We Tried" or "What Happened" unless it has a distinct outcome to report in the latter. Specifically:

• If an item appears in "What We Tried" (opened/created) and also merged or completed within the window, move it to "What Happened" (do not list in both).
• For items with multiple status transitions within the window, report only the most recent transition.
• Items that remain in the same state across both sections should appear only once, in whichever section is most relevant.

Phase 5: Post the Comment

Critical: Jira Cloud requires Atlassian Document Format (ADF) JSON for rich comment formatting. Markdown and Jira wiki markup both render as raw unformatted text. You must convert the comment to ADF before posting.

Converting to ADF

Build an ADF JSON document from the confirmed Markdown. The top-level structure is:

{
  "type": "doc",
  "version": 1,
  "content": [ ... ]
}

Map Markdown elements to ADF nodes:

Write the ADF JSON to a temp file (e.g., /tmp/pilot-update-adf.json).

Method 1 — acli (preferred)

acli is pre-authenticated in Ambient sessions (encrypted token stored in ~/.config/acli/jira_config.yaml). Post the comment with:

acli jira workitem comment create \
  --key PROJQUAY-11352 \
  --body-file /tmp/pilot-update-adf.json

Verify success by checking the exit code. If it fails, proceed to Method 2.

Method 2 — Jira REST API via curl (fallback)

This requires both JIRA_EMAIL and JIRA_API_TOKEN to be set (not available in all Ambient sessions). Source credentials from .claude/scripts/jira-ops.sh locations: JIRA_API_TOKEN env var (or ~/.config/acli/token.txt / ~/.acli-token), and JIRA_EMAIL (defaults to quay-devel@redhat.com if not set).

The ADF document must be wrapped in a body field. Set the default email if not already configured:

: "\${JIRA_EMAIL:=quay-devel@redhat.com}"

resp_file="$(mktemp)"
http_code="$(
  jq -n --argjson body "$(cat /tmp/pilot-update-adf.json)" '{body: $body}' | \
  curl -sS -o "$resp_file" -w "%{http_code}" -X POST \
    -u "${JIRA_EMAIL}:${JIRA_API_TOKEN}" \
    -H "Content-Type: application/json" \
    "https://redhat.atlassian.net/rest/api/3/issue/PROJQUAY-11352/comment" \
    -d @-
)"

if [[ "$http_code" != 2* ]]; then
  echo "ERROR: Jira comment POST failed (HTTP $http_code):"
  cat "$resp_file" >&2
fi

If both methods fail, save the ADF JSON to a file and provide the user with both options:

• A manual curl command they can run with their own credentials
• The original Markdown so they can format it manually in the Jira editor

Report the error.

Confirm success with: "Comment posted to PROJQUAY-11352."

Error Handling

• Repo not accessible: skip that repo, note it was skipped in the output.
• No pilot-relevant PRs found: include "No pilot-relevant PRs found in this period" in "What Happened".
• BFS finds no children under PROJQUAY-11352: report only PROJQUAY-11352 itself; this is valid.
• Jira batch changelog limit: if >20 keys, batch in groups of 20.
• User provides empty input for a subjective section: use "Nothing to report this period." as the placeholder.
• Comment post fails: save ADF JSON, provide a manual curl command and the original Markdown for Jira editor formatting. Report the error.
• Jira MCP tools unavailable (running outside Ambient): fall back to the Jira REST API via curl for both data gathering and comment posting. If credentials are not available, report the error and display instructions for manual posting.