backlog

name	backlog
description	Daily backlog management — full planning review OR add a single entry from a URL. Full review: gathers Jira sprint, GitHub PRs, checked-off items, and collaboratively plans the day. Add from URL: looks up a Teams message, Jira ticket, GitHub PR/issue, or incident.io alert and creates a concise backlog entry with a [ref] link. Triggers: /backlog-review, /update-backlog, "update my backlog", "plan my backlog", "what should I work on today", "refresh backlog", "backlog update", /backlog-add, "add this link to backlog", "backlog from link", "look up and add to backlog"
user_invocable	true

/backlog — Daily Backlog Management

Two modes: full daily review (default) or add single entry from URL (when a URL is provided).

Backlog Location

File: /Users/meain/.local/share/sbdb/Backlog/Backlog.md Sections: Today, Tomorrow, This Week, Weekend, Next Week, Whenever, Ongoing, and dated "Before" sections.

---

Mode A: Add Entry from URL

When given a URL, classify the source, fetch content, and add a task to the backlog.

Step 1: Classify

Pattern	Type
teams.microsoft.com/l/message/	Teams message
veeam-vdc.atlassian.net/browse/	Jira ticket
github.com/.../pull/	GitHub PR
github.com/.../issues/	GitHub issue
app.incident.io/	incident.io alert/incident

Step 2: Fetch Content

• Teams: Use MS365 MCP read_resource with URI teams:///teams/{groupId}/channels/{channelId}/messages/{messageId}. Extract sender name, message body.
• Jira: jira issue view PROJ-123 --plain — extract summary, status, assignee, type.
• GitHub PR: gh pr view <url> --json title,author,body,state
• GitHub issue: gh issue view <url> --json title,author,body,state
• incident.io: Use incident.io MCP tools (alert_show or incident_show).

Step 3: Compose Entry

Format: - [ ] <concise description> [ref](<original-url>)

Guidelines:

• Keep description short — a few words, enough to identify the task. No inline rationale or detail.
• If more context is needed, create a task note and link it with a wiki-link instead of expanding the description.
• Include person's first name if someone is asking/assigning something
• Jira: Fix auth token storage bug [DP-1234](url)
• GitHub PR: Review [repo#number](url): summary
• Teams: Review topic from Name [ref](url)
• Use [x] for completed or won't-do items. Never use [-] or any other checkbox state.

Step 4: Add to Backlog

Use /vault skill to add to the specified section (default: Today). If unavailable, directly edit Backlog.md — find the target section heading and insert at the end of that section.

---

Mode B: Full Daily Review

Collaboratively update the Today section of the backlog by pulling data from external sources and working with the user to decide what's realistic.

Step 0: Verify Today's Date

Always run date "+%Y-%m-%d (%A)" and treat its output as authoritative.

Monday cleanup: /Users/meain/.local/share/ignored-prs-for-review.yaml is a YAML file with three top-level list keys: prs (specific PR path fragments, e.g. owner/repo/pull/123), repos (bare repo names — blocks every PR from that repo), and people (author logins — blocks every PR by that person). Only the prs list resolves to real PRs, so Monday cleanup only touches that section — leave repos and people entries untouched.

If today is Monday, extract just the prs entries and check each one's state:

awk '/^prs:/{f=1;next} /^[a-zA-Z_]+:/{f=0} f' /Users/meain/.local/share/ignored-prs-for-review.yaml | sed -E 's/^[[:space:]]*- //'

For each non-empty result line:

export GIT_DIR=$(jj git root 2>/dev/null || echo .git)
result=$(gh pr view "https://github.com/$prpath" --json state 2>&1)
state=$(echo "$result" | grep -o 'OPEN\|MERGED\|CLOSED')

(use a variable name other than path — zsh aliases lowercase path to $PATH, and overwriting it breaks every subsequent command in the shell).

Remove any prs entries where state is not OPEN (merged, closed, or unknown/deleted), rewriting only that list in the YAML file — leave repos and people untouched. Report how many were pruned.

Step 1: Read Backlog

Read the full backlog file first.

Step 2: Handle Checked-Off Items

Note the checked-off items from Today — they will be archived to Before during Step 5. Format: ### YYYY-MM-DD (DayOfWeek). Get the date with:

~/.agents/skills/backlog/scripts/date-before.sh

Insert the new date block at the end (bottom) of the Before section — Before is chronologically ascending (oldest at top, newest at bottom).

Step 3: Gather External Data (parallel)

All Jira and GitHub commands must be run with dangerouslyDisableSandbox: true — the sandbox blocks TLS certificate verification needed for these hosts.

3a. Jira sprint:

jira issue list -q "sprint in openSprints() AND assignee = currentUser()" --raw 2>&1

Returns JSON array with raw Jira issue data. Parse with jq to extract key, status, summary.

Also fetch the sprint end date for the "This Cycle" section header:

jira sprint list --project DP | awk '/active/ {print $6}' | head -1

Returns a date like 2026-06-29. Use it in the section header as ### This Cycle (ends YYYY-MM-DD). If the backlog already has a "This Cycle" header with a different date, update it to reflect the current sprint.

3b. GitHub PRs — pending reviews:

~/.agents/skills/backlog/scripts/pending-prs.sh

Returns JSON array of PRs where user is a direct reviewer (not via team), excluding approved PRs, drafts, and Veeam-VDC/vdc-shared-data-plane PRs (never surface data-plane PRs — not the user's area). Also exclude dependabot PRs from Veeam-VDC repos — they are auto-reviewed and merged.

After fetching, filter out ignored PRs/repos/people. Flatten the YAML blocklist to one pattern per line, then filter the one-PR-per-line JSON with grep -vFf:

~/.agents/skills/backlog/scripts/pending-prs.sh | jq -c '.[]' | grep -vFf <(grep -E '^[[:space:]]*- ' /Users/meain/.local/share/ignored-prs-for-review.yaml | sed -E 's/^[[:space:]]*- //')

Apply the same flatten-and-filter pattern (jq per-line, then grep -vFf against the flattened YAML) whenever filtering PR lists in this skill, including the meain/* repos list in 3b-ii.

3b-ii. PRs from meain/ repos needing review:* Open PRs in repos owned by the meain GitHub org (user's personal projects where others open PRs):

GIT_DIR=$(jj git root 2>/dev/null || echo .git) gh search prs --state=open --owner=meain --json number,title,url,repository,updatedAt,isDraft,author 2>&1

Include all non-draft PRs not authored by meain. Filter out ignored PRs/repos/authors using the same flatten-and-filter pattern as 3b. Surface these in a separate "meain/* repos" subsection of the candidates file.

3c. PRs authored by user: Open PRs authored by the user across Veeam-VDC org.

3d. Calendar: Use MS365 MCP outlook_calendar_search to fetch today's meetings.

3e. This Cycle candidates: From the Jira sprint results (Step 3a), identify items that are not Done and not already present anywhere in the backlog file (check all sections — Today, Tomorrow, This Week, This Cycle, Next Week, Whenever). These are sprint commitments that have fallen off the radar.

Step 4: Select Candidates via Emacs

Write all candidates to /tmp/backlog-candidates-YYYY-MM-DD.md with - [ ] lines, grouped by:

• Morning routine (Apply rosemary water, Apply sunscreen, Check emails, Check Slack, Check Teams, Charge devices)
• Today's meetings (from calendar — list as info, no checkboxes; don't add to backlog)
• My PRs — approved (ready to merge)
• My PRs — needs action
• PRs needing direct review — split into subsections:
  • Recent (this week): PRs updated within the last ~7 days (only where login: meain is a direct reviewer, not team; exclude PRs with "local-dev" in the title, never include vdc-shared-data-plane PRs, and never include dependabot PRs from Veeam-VDC repos — they are auto-reviewed)
  • Older (sitting for a while): PRs more than ~7 days old that are still open and awaiting review (same exclusion rules apply) — useful for surfacing stale reviews that may have been overlooked
  • meain/ repos*: Open PRs in meain-owned repos (from Step 3b-ii), grouped separately — always include these regardless of age
• This Cycle — sprint items not yet in the backlog (from Step 3e); selected items go into the ### This Cycle section, not Today
• Candidates from other sections (Tomorrow/This Week/Whenever worth pulling in)

The "Checked-off items to move to Before" header must include the day name, e.g. (2026-06-03, Wednesday).

Inline Jira details: For any candidate line that references a Jira ticket (e.g. [DP-1234](url) or a DP-1234 in a PR title), fetch the ticket with jira issue view DP-1234 --plain (run in parallel for all tickets, dangerouslyDisableSandbox: true) and add a concise detail line immediately below the candidate, indented and prefixed with ↳. Format: ↳ DP-1234 (Type, Status, assignee/mine): one-line summary of the ask. This lets the reader glance at ticket context while selecting. These ↳ lines are candidate-file only — do NOT carry them into Backlog.md, which stays single-line per item.

Invoke the /edit-in-emacs skill with the file path (blocking, 600s timeout). After it returns, treat remaining - [ ] lines as selections.

Handling (ignored) markers: After the emacs edit returns, scan the candidates file for any lines containing (ignored). For each such line, extract the GitHub URL and append it (as a path fragment like owner/repo/pull/N) to /Users/meain/.local/share/ignored-prs-for-review. Do not add these PRs to the backlog.

Never use AskUserQuestion / multi-select prompts — always go straight to the emacs file.

Step 5: Commit Then Update Backlog

Before writing any changes to the backlog file, commit the current state:

cd /Users/meain/.local/share/sbdb && git add -A && git commit -m "Before backlog updates"

Then:

1. Move items from other sections to Today (don't duplicate)
2. Add new "This Cycle" candidates to the ### This Cycle section (not Today)
3. Add new items with reference links
4. Preserve priority emojis near top of Today

Step 6: Code Review Summaries (on request only)

Do not generate these automatically — the user rarely reads them. Only do this if explicitly asked (e.g. "generate summaries for today's reviews", "add a code review summary for cpb#5046").

When asked, spin off parallel background agents — one per requested PR. Each agent should:

1. Fetch the PR diff and details from GitHub
2. Run /explain-flow on the PR to generate a code flow summary
3. Write the result to:

   /Users/meain/.local/share/sbdb/Artifacts/Code Reviews/<org>/<repo>/<pr-number>.md
   

   For example: Artifacts/Code Reviews/Veeam-VDC/control-plane-backend/5046.md
4. Update the matching backlog line to append a markdown link:

   - [ ] Review [cpb#5046](https://github.com/...): title (author) [summary](Artifacts/Code%20Reviews/Veeam-VDC/control-plane-backend/5046.md)
   

   Use [summary] as link text. Path is vault-root-relative (no ../). URL-encode spaces (Code%20Reviews).
5. Commit the updated backlog file.

Formatting Rules

• Prefer inline links over trailing links (e.g. Merge [repo#123](url))
• Trailing [ref](url) only for URLs without a natural inline anchor (Teams links)
• Each PR must be on its own line
• Never remove section headers (Today, Tomorrow, etc.) even when empty
• Show meeting times in IST (UTC+5:30)
• Keep all backlog entries short — a few words per line. Use task notes for detail.
• Only two checkbox states: - [ ] (open) and - [x] (done/won't-do). Never use [-].