team-brief

# COG Team Intelligence Brief Skill

## When to Invoke
- User says "team brief", "team update", "what did we ship", "standup prep"
- User wants to know what the dev team is working on
- User needs cross-referenced GitHub + Linear + Slack intelligence
- Morning routine for product/engineering leads

## Agent Mode Awareness

**Check `agent_mode` in `00-inbox/MY-PROFILE.md` frontmatter:**
- If `agent_mode: team` — use the full parallel agent execution strategy below (6 agents)
- If `agent_mode: solo` — run data collection sequentially in the main conversation. Skip Phase 3.5 (Linear sync-back) and Phase 3.7 (HackMD publish) to keep it fast.

## Purpose
Generate a concise daily team intelligence brief focused on what we're doing, what needs attention, and what's falling through the cracks. Cross-references Linear work tracking, Slack discussions, GitHub activity, PostHog product analytics, braindumps, and recent meetings to surface insights, flag mismatches, and keep the team aligned.

**Monday = Week Start Brief:** If today is Monday, the brief covers Friday through Sunday (the full weekend window). All date ranges, queries, and language should reflect "since Friday" / "over the weekend" instead of "yesterday." This captures anything that happened Friday afternoon, weekend deploys, or async discussions.

**This is NOT an industry news report.** News/market intelligence belongs in the weekly brief.

## Command: `/team-brief`

## Voice & Tone

You're not a reporting tool — you're a teammate who happens to have read every Slack thread, every PR, and every meeting note. Write like someone who genuinely cares about the team succeeding.

**Do:**
- Be direct and opinionated. If something looks off, say it plainly.
- Celebrate wins. When the team ships something great, acknowledge it.
- Connect the dots. Don't just list facts — explain why they matter together.
- Use natural language. "We shipped v1.2.0 with 11 PRs — solid day" not "11 PRs were merged across the reporting period."
- Be concise but warm. Short sentences, clear thinking, no fluff.
- Flag risks honestly. "This worries me" or "worth keeping an eye on" — not corporate hedging.
- Show you understand the business context. Reference growth targets, customer feedback, strategy when relevant.

**Don't:**
- Write like a project management dashboard. No "REPORT_GENERATED_AT" energy.
- Be overly formal. No "It is recommended that the team consider..."
- Add filler. If there's nothing to say, don't pad it.
- Be passive. "The PR remains unreviewed" → "Nobody's reviewed this yet — it's been 3 days."
- Use unnecessary labels or tags. Let the writing speak for itself.

**Example tone:**
> Good day yesterday — we cut v1.2.0 with the new workflow, caching fix, and a bunch of UX polish. 11 PRs merged by 4 people. The team is shipping.
>
> But we have a problem: two discovery calls in a row ended with the prospect unable to log in. The onboarding flow just doesn't work. We're building features for users who can't get through the front door.

---

## Parallel Agent Team Execution Strategy

**The daily brief MUST use parallel agents to maximize speed and efficiency.**

When `/daily-brief` is invoked, the orchestrator (main agent) should:

### Phase 1: Setup (Orchestrator — Sequential, Fast)
1. Determine today's date and the **lookback start date**:
   - **If today is Monday:** lookback = last Friday (3 days ago). Set `LOOKBACK_DATE` to Friday's date. This is a **Week Start Brief**.
   - **Any other day:** lookback = yesterday. Set `LOOKBACK_DATE` to yesterday's date. This is a standard **Daily Brief**.
   - All dates in YYYY-MM-DD format.
2. Check available integrations via ToolSearch ("slack"), ToolSearch ("linear"), and ToolSearch ("posthog")
3. Check for meeting notes in `[CUSTOMIZE: path/to/meetings/]` matching dates from `LOOKBACK_DATE` through today
4. Check for recent braindumps in `[CUSTOMIZE: path/to/braindumps/]` matching dates from `LOOKBACK_DATE` through today

### Phase 2: Parallel Data Collection (Spawn 6 Agents Simultaneously)
**Launch ALL of these agents in a single message using the Task tool with `run_in_background: true`:**

#### Agent 1: "github-analyst" (subagent_type: general-purpose)
Prompt: see `references/agent-prompts.md` § github-analyst.

#### Agent 2: "slack-monitor" (subagent_type: general-purpose)
*Only spawn if Slack MCP is available*
Prompt: see `references/agent-prompts.md` § slack-monitor.

#### Agent 3: "meeting-reviewer" (subagent_type: general-purpose)
Prompt: see `references/agent-prompts.md` § meeting-reviewer.

#### Agent 4: "linear-tracker" (subagent_type: general-purpose)
*Only spawn if Linear MCP is available*
Prompt: see `references/agent-prompts.md` § linear-tracker.

#### Agent 5: "braindump-reviewer" (subagent_type: general-purpose)
Prompt: see `references/agent-prompts.md` § braindump-reviewer.

#### Agent 6: "posthog-analyst" (subagent_type: general-purpose)
*Only spawn if PostHog MCP is available*
Prompt: see `references/agent-prompts.md` § posthog-analyst.

### Phase 3: Assembly & Cross-Reference Synthesis (Orchestrator — After All Agents Complete)

This is the most critical phase. The orchestrator must:

1. **Collect results** from all background agents using Read on their output files

2. **Cross-reference Linear ↔ GitHub** (the core alignment check):
   - Match Linear issues to corresponding PRs/commits (by title, description, or explicit references)
   - Flag Linear issues marked "In Progress" with NO corresponding PR or commit (work claimed but not started in code)
   - Flag PRs that have NO corresponding Linear issue (untracked work — could be fine, just note it)
   - Flag Linear issues marked "Done" but PR still open or unmerged (status mismatch)
   - Note issues where Linear priority doesn't match PR urgency (e.g., Urgent in Linear but PR unreviewed for days)

3. **Cross-reference Slack ↔ GitHub ↔ Linear**:
   - Match Slack discussion topics to related PRs/commits AND Linear issues
   - Combine