Project Tracking

Finding Context

Use the context tool first to check what you already know:

# Check if an issue has been discussed before
context issue_history --issueId="ENG-123"

# See what came out of a previous meeting
context meeting_outcomes --videoId="vid-abc"

# Get context on people in the meeting
context person_context --email="[email protected]"

Then use Linear tools to fill gaps:

linear_project list              # See active projects
linear_issue list project:"X"    # Find existing tickets in a project
linear_issue list assignee:"Y"   # Find someone's current work

Creating Issues

For each new actionable item:

1. Check for duplicates — Search by keywords first
2. Create with context:
   • Clear title describing the deliverable
   • Description with relevant context
   • Assignee from the owner/action item
   • Project linkage
   • Priority based on urgency discussed

Title patterns:

• [Action] Verb + object — "Draft API spec for auth service"
• [Decision] What was decided — "Use PostgreSQL for user data"
• [Follow-up] Topic — "Revisit pricing model after beta"

Description template:

[Relevant context]

**Success criteria:**
- [ ] Criterion 1
- [ ] Criterion 2

Updating Existing Issues

When new info relates to tracked work:

1. Add comment with the update: linear_comment add
2. Update fields if needed: linear_issue update
   • State change (blocked → in progress)
   • Priority adjustment
   • Description additions

Issue Hierarchy

• Epics — Large initiatives, use as parent issues
• Sub-issues — Break down epics into deliverables
• Relations — blocks, blocked-by, related for dependencies

By Information Type

Meetings

Before processing a meeting, check context:

1. context person_context for each attendee — what are they working on?
2. Look for existing issues that might be discussed

Then extract:

• Action items → new issues with owners
• Decisions → comments on affected issues or decision records
• Status updates → state changes, comments
• Blockers → update issue state, add blocker notes

Correlations are tracked automatically — when you create/update issues while processing a meeting, the context plugin links them.

Meeting ↔ Issue Linking

When linking meetings to issues (from a button click or when creating tickets):

1. Search Linear for issues matching meeting topics, project names, and keywords from the summary
2. Present matches to the user for confirmation
3. For each linked issue, add a comment summarizing what was discussed about that issue in the meeting

When creating new tickets from a meeting:

• Check if any action items relate to existing issues (update those instead of creating duplicates)
• Fetch the meeting with bluedot_meeting get, extract action items from the summary yourself, then present for approval before creating

Messages/Requests

• Direct asks → new issues if actionable
• Questions about work → find and share relevant issues
• FYIs → comment on related issues if context-adding

Status Reports

• Progress → update issue states
• Risks → add to descriptions or create blocker issues
• Completions → move to done, add summary comment

Approval Workflow

When you identify actions to take, present them for approval first:

1. Summarize what you learned
2. List proposed actions:
   • Issues to create (with title, assignee, project)
   • Issues to update (with what changes)
   • Comments to add
3. Wait for user confirmation before executing

Example:

From this meeting, I'd propose:

**Create:**
- "Implement OAuth flow" → @alice, Auth project, P2
- "Update API docs" → @bob, Docs project, P3

**Update:**
- ENG-123: add comment about new timeline
- ENG-456: change state to blocked

Proceed?

Once approved, execute all actions and confirm completion.

Anti-patterns

• Creating duplicate issues without searching first
• Tickets without clear owners or deliverables
• Overly granular tickets (every sentence becomes an issue)
• Forgetting to link related work
• Stale tickets — update or close, don't ignore
• Attributing information to people ("James said...", "From meeting...")
• Source metadata in comments — just state the information directly