Opportunity Scanner

Deep-dive your connected work tools to find the highest-value Claude skills, projects, and artifacts you should build — for yourself and for your team.

This isn't a survey. It scans your actual work patterns, finds the waste, and tells you exactly what to build and why.

When to Use

Invoke when user:

• Wants to know what Claude skills or projects would save them the most time
• Says "what should I build?" or "scan for opportunities"
• Is a leader looking for AI wins across their department
• Wants ROI-backed recommendations, not guesses
• Has MCPs connected and wants to put them to work strategically

Pre-Flight Check

Before scanning, detect what tools are connected. Probe each MCP silently — don't ask the user to list them.

Check for availability of:

• Google Calendar (list events)
• Gmail (search messages)
• Slack (search/read channels)
• Jira (search issues)
• Confluence (search pages)
• Google Docs/Sheets (search/list)

Report what's connected:

Connected tools detected:

• ✅ Google Calendar
• ✅ Gmail
• ✅ Slack
• ❌ Jira (not connected)
• ❌ Confluence (not connected)

I'll scan everything that's connected. The more tools available, the better the recommendations. Ready to start?

If fewer than 2 tools are connected, warn that recommendations will be limited but proceed anyway.

Core Workflow

Phase 1: Deep Scan (3-5 minutes)

Scan all connected sources in parallel. Pull data from the last 30 days unless the user specifies otherwise.

Calendar Scan

Pull: All events from the last 30 days Analyze for:

• Meeting load: Total hours/week in meetings. Trend over 4 weeks.
• Recurring meetings: Which ones repeat? How many attendees? Duration?
• 1:1 density: How many 1:1s per week? With whom?
• Meeting types: Classify as status update, decision-making, brainstorm, 1:1, all-hands, external
• Prep patterns: Are there prep blocks before meetings? Gaps that suggest context-switching?
• Meeting-free time: How many focused hours per week?

Opportunity signals:

• Status meetings > 3/week → could be async briefings
• Recurring meetings with 5+ people → could have AI-generated pre-reads
• 1:1s with direct reports → prep could be automated
• Back-to-back blocks with no prep time → meeting prep skill needed

Email Scan

Pull: Last 200 emails (sent + received), search for high-frequency patterns Analyze for:

• Volume: Emails sent/received per day
• Response patterns: Average response time. Who do they respond to fastest/slowest?
• Repetitive content: Similar emails sent to different people (templates hiding in plain sight)
• Request patterns: What do people ask them for repeatedly?
• FYI vs action: Ratio of informational emails to ones requiring action
• Thread length: Long threads suggest decisions that could be structured differently

Opportunity signals:

• Same type of email sent 3+ times/month → template or auto-draft skill
• Long response times to certain senders → triage/prioritization needed
• High FYI volume → digest/summary skill
• Repeated requests for the same info → knowledge base or self-serve artifact

Slack Scan

Pull: Channels they're active in, messages from last 30 days, DMs (last 14 days) Analyze for:

• Channel load: How many channels? Active in how many?
• Message volume: Messages sent/received per day
• Question patterns: What do people ask them in DMs? In channels?
• Repeated questions: Same question from different people → documentation gap
• Status update patterns: Are they manually posting updates that could be automated?
• Information routing: Are they the bridge between teams? Forwarding/translating between groups?
• @mention response time: How fast do they respond to direct mentions?
• Thread patterns: Do they start threads or respond? Long threads suggest complex decisions.

Opportunity signals:

• Same question asked 3+ times → FAQ artifact, knowledge skill, or Confluence page
• Manual status updates → automated status report project
• High DM volume with repeated patterns → team communication skill
• Bridge between 2+ teams → cross-functional digest or routing automation
• Slow @mention response → triage/notification skill needed

Jira Scan (if connected)

Pull: Issues assigned to them, issues they've created, recent activity Analyze for:

• Ticket patterns: What types of tickets do they create/handle most?
• Status transitions: How long do tickets sit in each state?
• Comment patterns: Repetitive comments or status updates?
• Sprint patterns: Consistent velocity? Spillover?
• Blocker frequency: What blocks work most often?

Opportunity signals:

• Repetitive ticket creation → ticket template or auto-creation skill
• Long time in "waiting" states → escalation or follow-up automation
• Manual status comments → automated status updates
• Recurring blocker patterns → proactive blocker detection

Confluence Scan (if connected)

Pull: Pages they've authored/edited recently, search for team documentation Analyze for:

• Documentation freshness: When were key docs last updated?
• Page creation patterns: What kind of docs do they create?
• Gap analysis: Topics discussed in Slack/email that have no Confluence page

Opportunity signals:

• Stale documentation → doc refresh project
• Topics discussed repeatedly without docs → documentation generation skill
• Frequently referenced pages → could become interactive artifacts

Google Docs/Sheets Scan (if connected)

Pull: Recently accessed/created documents Analyze for:

• Document types: Reports, plans, templates, meeting notes?
• Creation frequency: How often do they create similar docs?
• Collaboration patterns: Shared with whom? Commented on by whom?

Opportunity signals:

• Similar documents created repeatedly → template or generation skill
• Manual data compilation in sheets → automated dashboard artifact
• Meeting notes for every recurring meeting → auto-notes skill

Phase 2: Pattern Analysis (1-2 minutes)

After scanning, synthesize findings across all sources. Look for cross-source patterns:

Cross-Source Pattern Detection

ROI Scoring

For each identified opportunity, calculate:

Weekly Time Saved = frequency_per_week × time_per_instance × automation_percentage
Annual ROI Hours = Weekly Time Saved × 48 weeks
Impact Score = (Annual ROI Hours × people_affected) / implementation_effort

Frequency: How often does this pattern occur? (from scan data) Time per instance: Estimated minutes per occurrence Automation percentage: How much can Claude realistically handle? (be honest — 40-80%, not 100%) People affected: Just them, or their whole team? Implementation effort: Low (1 hour), Medium (half day), High (full day)

Phase 3: Recommendation Generation

Generate recommendations in two categories. Each recommendation must include:

• What to build (skill, project, or artifact — be specific)
• Why (the pattern that triggered this recommendation, with data)
• ROI estimate (hours saved per week, who benefits, implementation effort)
• How (2-3 sentence description of what the build looks like)

Category 1: For You (Personal Productivity)

Recommendations that help THIS person work faster, stay on top of things, reduce cognitive load.

Prioritize by:

1. Highest weekly time savings
2. Lowest implementation effort
3. Daily pain points over weekly ones

Types to recommend:

Category 2: For Your Team (Department Leadership)

Recommendations that help their TEAM or DEPARTMENT. These are force-multiplier plays.

Prioritize by:

1. Number of people who benefit
2. Eliminates repeated friction (same question, same manual process)
3. Reduces dependency on any single person (especially them)

Types to recommend:

Phase 4: Deliver Report

Present findings in a structured, scannable format.

# Opportunity Scan Report — [Name], [Role]
**Scan date**: [Date]
**Sources scanned**: [List connected tools]
**Time window**: Last 30 days

---

## Key Findings

**Your work profile**:
- [X] hours/week in meetings ([trend] from last month)
- [X] emails/day ([X] requiring action, [X] FYI)
- [X] Slack channels active, [X] DMs/day
- [X] Jira tickets touched/week

**Top patterns detected**:
1. [Pattern name] — [one-line description with data]
2. [Pattern name] — [one-line description with data]
3. [Pattern name] — [one-line description with data]

---

## For You: Top [N] Recommendations

### 1. [Build Name] — [Skill/Project/Artifact]
**ROI**: ~[X] hours saved/week | Effort: [Low/Med/High]
**Pattern**: [What we found in the data]
**Build**: [2-3 sentences on what this looks like]

### 2. ...

---

## For Your Team: Top [N] Recommendations

### 1. [Build Name] — [Skill/Project/Artifact]
**ROI**: ~[X] hours saved/week across [N] people | Effort: [Low/Med/High]
**Pattern**: [What we found in the data]
**Build**: [2-3 sentences on what this looks like]

### 2. ...

---

## Quick Wins (< 1 hour to build)
- [ ] [Quick win 1] — [one line]
- [ ] [Quick win 2] — [one line]
- [ ] [Quick win 3] — [one line]

---

## Implementation Roadmap

| Priority | Build | Type | Effort | ROI/week | Start |
|----------|-------|------|--------|----------|-------|
| 1 | [Name] | Skill | Low | 3h | Now |
| 2 | [Name] | Project | Med | 5h | This week |
| 3 | [Name] | Artifact | Low | 2h | Now |
| ... | ... | ... | ... | ... | ... |

Phase 5: Interactive Deep-Dive

After delivering the report:

That's the scan. Want to:

1. Deep-dive a specific recommendation (I'll spec it out in detail)
2. Build one now (I'll generate the skill, project instructions, or artifact)
3. Rescan with a different time window or focus area
4. Export the full report as a document

If they choose to build, generate the complete output:

• Skill: Full SKILL.md with references
• Project: Complete custom instructions ready to paste
• Artifact: The actual artifact (dashboard, template, FAQ, etc.)

Presentation Rules

• Lead with data, not opinions. Every recommendation cites scan evidence.
• Be honest about ROI. Don't inflate numbers. 2 hours/week saved is meaningful — don't round up to 5.
• Separate personal from team. Leaders think in both frames. Don't mix them.
• Quick wins first. Low-effort, high-frequency wins build momentum.
• No jargon about AI. Don't say "leverage AI to optimize workflows." Say "this scans your Jira every morning and posts a status update so you don't have to."
• Specific > generic. "Build a meeting prep skill for your weekly product review" beats "Consider automating meeting preparation."

Anti-Patterns

• Don't recommend things the user clearly doesn't need. If they have 3 meetings a week, don't recommend a meeting prep skill.
• Don't assume all tools are connected. Only recommend based on what's actually available.
• Don't recommend building what already exists in the environment. Check if the skill/project already exists before recommending.
• Don't overload the report. Max 5 personal + 5 team recommendations. Quality over quantity.
• Don't hand-wave implementation. Every recommendation should be buildable. If it's not, don't recommend it.
• Don't scan private/sensitive content beyond what's needed. Summarize patterns, don't quote private messages in the report.

Privacy & Sensitivity

• Never quote private DMs or emails verbatim in the report. Summarize patterns only.
• Don't name individuals in negative contexts. "Your team asks the same onboarding question frequently" not "Sarah asked the same question 4 times."
• Aggregate, don't expose. Team recommendations should reference patterns, not individuals.
• Skip HR/legal/compensation signals. If the scan surfaces sensitive topics, ignore them — they're not automation candidates.

References

For detailed protocols:

• references/scan-protocols.md — MCP query patterns for each data source
• references/pattern-catalog.md — Full catalog of cross-source patterns and their signals
• references/roi-framework.md — Detailed ROI calculation methodology
• references/recommendation-templates.md — Example recommendations by role and industry