Quick Reference & Central Insights

This lesson is a reference. There are no new concepts here — only clear tables and the central insight that ties the chapter together. Return here whenever you need to remember which command handles which task, which plugin provides it, or which configuration field controls a specific output.

The Central Insight

Explicit knowledge becomes findable without asking a person. Tacit knowledge gets captured before it walks out the door. HR professionals are freed for the 40% that only humans can do.

This is what Chapter 37 builds toward. The 14 commands and 4 agents in this chapter do not replace HR judgment — they eliminate the administrative overhead that crowds it out. Every command in the reference below handles a task that belongs to the 60%. Every escalation path in every lesson points to the 40% that stays with humans.

Complete Command Reference

Official Plugin: human-resources (Anthropic)

Install: Cowork → Customize → Browse plugins → "Human Resources"

Command	Function	Sensitivity	Covered In
/policy-lookup	Find and explain policies in plain language	ROUTINE	Lesson 3
/onboarding	Onboarding plans and schedules	ROUTINE	Lesson 5
/draft-offer	Offer letters and employment documents	CONFIDENTIAL	Lesson 7
/interview-prep	Interview questions and rubrics	ROUTINE	Lesson 6
/performance-review	Performance reviews	CONFIDENTIAL	Lesson 8
/comp-analysis	Compensation analysis and benchmarking	CONFIDENTIAL	Lesson 9
/org-planning	Organisational planning and restructuring	CONFIDENTIAL	Lesson 9
/people-report	People analytics and reporting	CONFIDENTIAL	Lesson 13
/recruiting-pipeline	Recruiting pipeline management	CONFIDENTIAL	Lesson 13

Custom Plugin: hr-operations (Panaversity)

Install: Cowork → Customize → + → Add marketplace from GitHub → panaversity/agentfactory-business-plugins → hr-operations

Command	Function	Sensitivity	Covered In
/jd	Job descriptions with inclusive language	ROUTINE	Lesson 6
/match	Internal talent matching and succession	CONFIDENTIAL	Lesson 9
/knowledge	Institutional knowledge capture	CONFIDENTIAL	Lesson 10
/reference	Reference letters and employment verification	CONFIDENTIAL	Lesson 7
/offboard	Offboarding processes and knowledge transfer	CONFIDENTIAL	Lesson 11

Agent Summary

All four agents are part of the custom hr-operations plugin. Each operates as a persistent background process — triggered automatically, not by manual command invocation.

Agent	Purpose	Trigger	Schedule	Sensitivity
knowledge-base-agent	24/7 employee Q&A with escalation	Always-on (Slack/Teams)	Continuous	ROUTINE (queries)
onboarding-orchestrator	New hire workflow automation (T-14 to Day 90)	HRIS new hire record	T-14 through Day 90	CONFIDENTIAL
policy-maintenance-agent	Policy currency and statutory rate monitoring	Monthly + rate change events	1st Monday of each month	ROUTINE
offboarding-knowledge-agent	Departure knowledge capture	HRIS resignation record	Within 24 hours of trigger	CONFIDENTIAL

Agent setup, configuration, and deployment are covered in Lessons 4, 5, 11, and 12.

Sensitivity Classification Guide

Every AI output in this chapter carries one of three sensitivity labels. The label tells you how to handle the output — who can see it, how to store it, and what to do if something goes wrong.

Label	When to Use	Examples	Handling
ROUTINE	Policy summaries, JDs, general queries, onboarding plans	/policy-lookup output, /jd output, knowledge-base-agent general responses	Standard handling — can be shared with employees
CONFIDENTIAL	Anything containing personal data, salary, performance, or talent information	/draft-offer, /performance-review, /match, /reference, /comp-analysis	Contains personal/confidential data — handle per your data protection policy
SENSITIVE PERSONAL DATA	Medical, disciplinary, grievance, termination	Any output touching health, conduct investigations, or dismissal	NEVER auto-generate — escalate immediately to a named HR contact

The Escalation Rule

When any query touches SENSITIVE PERSONAL DATA territory, the correct response is always a warm handoff:

I can see this involves [medical / disciplinary / grievance] information.
This is outside what I can assist with directly. Please contact
[named HR contact] at [email / phone] who can handle this with
the appropriate care and confidentiality.

This pattern appears in every lesson. It is not a limitation of the tools — it is the boundary between what AI can do and what only humans should do.

hr.local.md Configuration Reference

hr.local.md has eight sections. This table summarises what each section controls in plugin output.

Section	Fields	Controls
Organisation Profile	Name, size, sector, location, HRIS, HR team	Organisation identity in all outputs
Jurisdiction	Employment law jurisdiction, leave entitlement, notice periods, statutory sick pay	Employment rights figures and statutory references
Policy Library	Document locations for each policy	Source citations in /policy-lookup output
Benefits	Leave days, sick pay, parental leave, pension, health insurance, L&D budget	Benefits references in offer letters, onboarding, policy summaries
HR Contacts	Named contacts for HR, payroll, benefits, IT, legal, OHW, EAP	Escalation paths in all outputs
Onboarding Configuration	Pre-boarding window, 30/60/90 goals, buddy programme, systems to provision	/onboarding output structure and timelines
Performance Review Cycle	Frequency, months, rating framework, competency framework, 360° process	/performance-review structure and language
Reference Letters	Reference policy, authorised signatories, what is and is not confirmed	/reference output scope and compliance

Updating Statutory Rates

Statutory rates change annually in most jurisdictions (typically April in the UK; varies for Pakistan provincial minimum wages). When rates change:

1. Update the Jurisdiction section of hr.local.md with the new rate and the effective date
2. Verify by running /policy-lookup statutory sick pay and checking the figure matches
3. If the policy-maintenance-agent is running, it will flag the discrepancy automatically — but the update to hr.local.md is still manual

Adding New Policies

When your organisation adds or significantly revises a policy:

1. Add the new policy location to the Policy Library section
2. Run /policy-lookup [new policy topic] to verify the output references the correct document
3. If the policy-maintenance-agent is running, schedule a policy review in the agent's configuration to include the new policy in its next sweep

Chapter Contract Answers

By the end of Chapter 37, you should be able to answer all five of these questions. These were set in the Chapter README as the chapter's completion criteria.

1. What is the difference between explicit and tacit knowledge in HR — and why does the distinction matter?

Explicit knowledge exists in documents — theoretically findable. Tacit knowledge exists in people's heads — invisible until they leave. The distinction matters because each requires a different solution: explicit knowledge needs better findability (the knowledge-base-agent); tacit knowledge needs systematic capture before it disappears (/knowledge and the offboarding-knowledge-agent).

2. Which HR tasks belong to the 60% (AI can handle) and which belong to the 40% (humans must handle)?

The 60% is information routing, repetitive process execution, and structured knowledge capture — anywhere the answer follows a known pattern. The 40% is difficult conversations, sensitive investigations, genuine talent judgement, ethical decisions, and any situation touching individual dignity, health, or conduct. The boundary is the SENSITIVE PERSONAL DATA label — when a query reaches that territory, it belongs in the 40%.

3. How do you install and configure both HR plugins, and how does hr.local.md make outputs organisation-specific?

Install the official human-resources plugin via the Cowork store. Install the custom hr-operations plugin via Cowork → Customize → + → Add marketplace from GitHub → panaversity/agentfactory-business-plugins. Configure hr.local.md with the eight sections: organisation profile, jurisdiction, policy library, benefits, HR contacts, onboarding configuration, performance review cycle, and reference letter policy. Without hr.local.md, outputs use generic defaults. With it, every output references your actual entitlements, contacts, and policies.

4. What are the four persistent agents in the hr-operations plugin — what does each do and what triggers each?

knowledge-base-agent — always-on employee Q&A, triggered by Slack/Teams messages. onboarding-orchestrator — new hire workflow automation from T-14 to Day 90, triggered by HRIS new hire record. policy-maintenance-agent — policy currency and statutory rate monitoring, triggered monthly and by rate change events. offboarding-knowledge-agent — departure knowledge capture, triggered within 24 hours of an HRIS resignation record.

5. How do you apply the sensitivity classification (ROUTINE / CONFIDENTIAL / SENSITIVE PERSONAL DATA) to HR AI outputs — and what is the escalation rule?

ROUTINE outputs (policy summaries, JDs, general queries) can be shared with employees. CONFIDENTIAL outputs (offer letters, performance reviews, salary data, talent assessments) contain personal data — handle per your data protection policy. SENSITIVE PERSONAL DATA (medical, disciplinary, grievance, termination) must never be auto-generated — always escalate to a named HR contact with a warm handoff message.

What You Can Now Do

After completing Chapter 37, you can:

• Answer HR questions automatically using the knowledge-base-agent and /policy-lookup
• Produce HR documents in minutes — onboarding plans, job descriptions, offer letters, performance reviews, reference letters
• Capture institutional knowledge systematically before it walks out the door with departing employees
• Run four persistent agents that handle HR workflows automatically in the background
• Apply the sensitivity framework consistently — knowing when AI can help and when to escalate

What does not change: the need for human judgment in difficult situations. Terminations, grievances, medical situations, team conflicts, and ethical decisions require HR professionals who can exercise judgement, hold difficult conversations, and be present in ways that AI cannot be. Chapter 37 does not reduce the importance of that work — it creates the capacity to do it properly.

Flashcards Study Aid