jamesc/whats-next

What's Next Workflow

When activated, execute this workflow to recommend the next issue to work on:

Steps

1. Check current branch: Determine if we're on a feature branch and extract the issue number if present.

2. Check for active cycles: Query Linear for the BT team's active cycle:

   streamlinear-cli graphql "query { team(id: \"BT\") { key activeCycle { id name issues { nodes { id identifier title state { name } priority } } } } }"
   

   If an active cycle exists, prioritize issues from that cycle.

3. Get recent completed work: Query Linear for recently completed issues to understand work patterns:

   streamlinear-cli graphql "query { issues(filter: {team: {key: {eq: \"BT\"}}, state: {name: {in: [\"Done\", \"In Review\"]}}}, orderBy: updatedAt, first: 10) { nodes { id identifier title labels { nodes { name } } parent { identifier title } } } }"
   

4. Check git history: Review recent commits to understand what areas were recently worked on:

   git log --oneline -20 --format="%s" | grep -oE "BT-[0-9]+" | sort -u | head -5
   

5. Find candidate issues: Query Linear for backlog/todo issues that are ready to work on. Skip "In Progress" issues — those are already being worked on by another agent.

   streamlinear-cli graphql "query { issues(filter: {team: {key: {eq: \"BT\"}}, state: {name: {in: [\"Backlog\", \"Todo\"]}}}, orderBy: priority, first: 20) { nodes { id identifier title description priority state { name } labels { nodes { name } } parent { identifier title } children { nodes { identifier state { name } } } relations { nodes { type relatedIssue { identifier state { name } } } } } } }"
   

6. Prioritize issues: Score and rank issues based on:

   • Active cycle membership (highest priority if cycle is active)
   • agent-ready label (fully specified, can start immediately)
   • Priority level (1=Urgent, 2=High, 3=Medium, 4=Low)
   • Blocking relationships (issues that unblock others get priority)
   • Relatedness to recent work (same parent issue, similar area)
   • Dependencies satisfied (all blocking issues are Done)

7. Present recommendations: Display the top 3-5 recommended issues in this format:

   ## Recommended Next Issues
   
   | # | Issue | Title | Priority | Score | Why |
   |---|-------|-------|----------|-------|-----|
   | 1 | BT-XX | Title | High | 85 | In active cycle, unblocks BT-YY |
   | 2 | BT-ZZ | Title | Medium | 65 | agent-ready, related to recent BT-WW |
   | 3 | BT-AA | Title | Medium | 50 | agent-ready, no blockers |
   
   **Top pick: BT-XX** — [one sentence explaining why this is the best choice]
   

   For each issue include:

   • Issue identifier and title
   • Priority and score
   • Why it's recommended (cycle, related to recent work, unblocks others, etc.)
   • Any caveats (e.g., "large scope, consider breaking down")

8. Optional: Start work: Ask the user if they want to start on the top recommended issue (which would run the /pick-issue workflow for that specific issue).

Scoring Logic

| Condition | Points | |-----------|--------| | In active cycle | +50 | | Has agent-ready label | +30 | | Blocker completed in last 7 days (newly unblocked) | +25 | | Related to recently completed work (same parent or area) | +20 | | Unblocks other issues | +15 | | Priority 1 (Urgent) | +10 | | Priority 2 (High) | +5 | | Has unresolved blocking dependencies | -100 | | Has needs-spec label (requires human input first) | -50 | | Has blocked label | -20 | | Is "In Progress" (another agent working on it) | -200 |