Prove Professional Value

Everything you have built so far — identity, skill, connections, scheduling, boundaries, memory — converges here. This final project milestone has two tracks: Silver proves your employee delivers real professional value through autonomous reporting. Gold proves you can architect a multi-context system with proper isolation.

Both tracks test the same question: would you trust this employee's output enough to act on it? A report you would not show your boss is not a report worth generating. An architecture without real isolation is not an architecture worth building.

Choose your tier, or complete Silver first and continue to Gold.

---

Silver Deliverable: Autonomous Report

Build a report your employee generates and delivers on schedule. The report must do more than summarize — it must synthesize data from multiple sources and make at least one recommendation you did not ask for.

Silver Success Criteria

1. The report runs on schedule and is delivered through your configured channel
2. It reads from 2 or more data sources (files, APIs, databases, web content — whatever fits your domain)
3. It contains at least one proactive recommendation — something the employee spotted that you did not explicitly ask about. Drawing on patterns from your action log or knowledge store strengthens this significantly
4. It meets professional quality standards: you would share this with your boss, client, or colleague without embarrassment

Silver Use Case Gallery

Accountant — Weekly cash flow report: reads bank transactions and outstanding invoices, cross-references payment timelines, flags potential cash crunch 2 weeks before it hits, recommends which invoices to follow up on first.

Teacher — Weekly class progress report: reads assignment submission data and recent test scores, identifies students whose performance dropped more than 15% in the last two weeks, recommends targeted intervention strategies per student.

Consultant — Client health dashboard: reads project milestone status and client communication frequency, flags accounts where milestone delivery is on track but communication has dropped (early warning of silent dissatisfaction), recommends proactive check-in schedule.

Recruiter — Pipeline report: reads application volumes and interview outcome data, identifies stages where candidates disproportionately drop out, recommends sourcing adjustments or process changes to improve conversion.

---

Gold Deliverable: Multi-Group Architecture

Design a system where different professional contexts operate with different identities, permissions, and data access — all within one NanoClaw installation.

Gold Success Criteria

All Silver criteria, plus:

1. Three groups configured with distinct CLAUDE.md identities: main (admin and control), a professional work group (domain tasks), and a client-facing group (limited scope, restricted data access)
2. Demonstrated isolation: the non-admin group cannot perform admin actions or access admin-only data
3. Data boundaries documented and tested: clear rules for what data flows between groups and what stays isolated
4. A system architecture diagram showing groups, data flows, and permission boundaries
5. Memory isolation verified: each group's action log and knowledge store are independent — the client-facing group cannot query the admin group's action log or knowledge entries

Gold Use Case Gallery

Accountant — main (admin: system config, all data access) + accounting-ops (bookkeeping, report generation, full transaction data) + client-portal (read-only financial summaries, no raw transaction access, no ability to modify records). client-portal has its own knowledge store — corrections about Client A's preferences never leak to Client B's group.

Consultant — main (admin: system config, all project data) + research (full data access, deep analysis, internal memos) + client-comms (filtered output only, no internal strategy documents, no access to other client data). client-comms cannot access the internal group's action log — prevents leaking which other clients are being served.

Recruiter — main (admin: pipeline management, all candidate data) + sourcing (job descriptions, public candidate info, outreach templates) + hiring-manager-view (anonymized candidate summaries, interview feedback forms, no salary data)

Teacher — main (admin: all student records, system config) + curriculum (lesson planning, assessment design, aggregate performance data) + parent-portal (individual student progress only, no class-wide data, no internal notes)

---

Hints

Level 1 — Think about data sources and audience

For Silver: What two data sources, when cross-referenced, would give you the most valuable insight in your profession? The power of synthesis is in the combination — bank data alone is accounting, but bank data plus outstanding invoices is cash flow forecasting.

For Gold: Think about who interacts with your work. You have internal operations, client-facing outputs, and administrative control. Each of those contexts needs different data access and different behavior. Who should never see what?

Level 2 — Ask your AI to help design the system

For Silver:

Design a weekly [your profession] report that combines [data source A]
and [data source B]. The report should:
1. Summarize key metrics from each source
2. Cross-reference the two sources to find patterns or anomalies
3. Generate at least one proactive recommendation
4. Format the output for delivery via [your channel]

Include the specific data fields to read from each source and the
logic for generating recommendations.

Your action log from Give Your Employee a Memory is a natural data source. A daily digest of autonomous actions — "here's everything I did today without asking you" — is the simplest Silver report that proves real professional value. Combine it with one external data source for the 2-source requirement.

For Gold:

Design a 3-group NanoClaw architecture for a [your profession]:
- Group 1 (main): admin and system control
- Group 2 ([work-context]): day-to-day professional tasks
- Group 3 ([external-context]): limited external-facing access

For each group, specify:
1. The CLAUDE.md identity and personality
2. What data it CAN access
3. What data it CANNOT access
4. What actions it CAN perform
5. What actions are blocked

Then describe how to test that isolation actually works.

Level 3 — Step-by-step implementation guide

Silver implementation:

1. Create a skill that performs your report logic: (a) read source A, (b) read source B, (c) cross-reference for patterns — query your action_log for recurring patterns or your knowledge store for accumulated preferences, (d) generate recommendations, (e) format the report, (f) deliver via your channel
2. Configure the scheduler to run this skill on your desired cadence (e.g., every Monday at 8am)
3. Run it manually first and review the output. Iterate on the skill until the report quality meets your standard
4. Enable the schedule and let it run autonomously. Save the output as domain-report-sample.md

Gold implementation:

1. Create three group directories: groups/main/, groups/[work-group]/, groups/[external-group]/
2. Write a distinct CLAUDE.md for each group with different identity, permitted actions, and data access rules
3. Test isolation: from the non-admin group, attempt an admin action (like modifying system config). Verify it fails or is refused
4. Test data boundaries: from the external group, attempt to query the admin group's action_log or knowledge store. Verify the data is not available. This proves your memory isolation works at the group level, not just the CLAUDE.md level
5. Create a system architecture diagram showing the three groups, their data access, and the boundaries between them
6. Document all test results