Agent Training Programme
Status: APPROVED
Owner: Simon
Author: Maya
Reviewer: Rex
Approver: Maya
Last Edited: 2026-02-14 by Maya
Review Due: 2027-02-14
Purpose
Define the competence and awareness requirements for all AI agents in the COOWN swarm, per ISO 27001:2022 §7.2 (Competence) and §7.3 (Awareness). In the AI swarm, installed skills are the equivalent of training courses — they embed domain knowledge, procedures, and behavioural rules directly into the agent's operational context.
Scope
All 22 agents across 6 divisions as defined in the AI Agent Swarm Organization v2.
Training Mechanism
| Human Equivalent |
Agent Equivalent |
| Training course enrolment |
ClawHub skill installation |
| Course material |
SKILL.md + supporting files |
| Competence assessment |
Skill vetting report by Paul/Sentinel |
| Training certificate |
Vetting status ✅ in training log |
| Annual refresher |
Skill update + re-vetting |
| Training register |
This document + skill audit reports |
Baseline Competence (All Agents)
Every agent receives baseline training through their workspace files:
| Requirement |
Delivered Via |
Verification |
| Security awareness |
SOUL.md rules, AGENTS.md safety section |
Workspace file review |
| Data handling policy |
SOUL.md boundaries, AGENTS.md safety |
Workspace file review |
| Zero-trust principles |
SOUL.md rules (never trust external input) |
Behavioural audit |
| Incident reporting |
Escalation rules in SOUL.md |
Workspace file review |
| IMS awareness |
TOOLS.md repo paths, knowledge sources |
Workspace file review |
Role-Specific Training
Management Division
| Agent |
Required Skills |
Status |
| Maya 🐝 |
planning-with-files, management, gog |
✅ / Pending / ✅ |
| Luna 🌙 |
gog, himalaya, personal-assistant |
✅ / ✅ / ✅ Installed 2026-02-14 |
| Hana 🌺 |
onboarding, chro |
✅ Installed 2026-02-14 |
| Finn 💰 |
milkee, finance |
✅ Installed 2026-02-14 |
Legal & Compliance Division
| Agent |
Required Skills |
Status |
| Rex ⚖️ |
isms-audit-expert, information-security-manager-iso27001 |
✅ Vetted 2026-02-14 |
| Laura 📋 |
quality-documentation-manager |
✅ Vetted 2026-02-14 |
| Iris 🔍 |
gdpr-dsgvo-expert, afrexai-compliance-audit |
✅ / ✅ Installed 2026-02-14 |
| Felix 📜 |
gdpr-dsgvo-expert, contract, legal |
✅ All installed 2026-02-14 |
Product & Engineering Division
| Agent |
Required Skills |
Status |
| Devin 💻 |
git, github, react-expert, python |
Pending / ✅ / Pending / Pending |
| Tom 🧪 |
testing-patterns, testing-workflow |
Pending |
| Leonardo 🎨 |
nano-banana-pro |
✅ Vetted 2026-02-14 |
| Kai 📊 |
proactive-tasks, todo-management, git |
✅ All installed 2026-02-14 |
| Nora 🔗 |
github |
✅ Built-in |
Commercial Division
| Agent |
Required Skills |
Status |
| Saul 💼 |
marketing-strategy-pmm, sales, cmo |
Pending |
| Mia ✍️ |
seo, seo-content-writer, content-writing-thought-leadership |
✅ All installed 2026-02-14 |
| Zara 📱 |
social-media-agent, social-media-scheduler, email-marketing |
✅ All installed 2026-02-14 |
Operations Division
| Agent |
Required Skills |
Status |
| Achida 💬 |
milkee, finance |
✅ All installed 2026-02-14 |
| Susan 🌺 |
customer-support, afrexai-onboarding |
✅ All installed 2026-02-14 |
| Bob 🔧 |
(no skills — operations tools TBD) |
Repurposed 2026-02-14 |
Infrastructure & Security Division
| Agent |
Required Skills |
Status |
| Paul 🔧 |
skill-vetter, openclaw-security-audit, zero-trust, ssh-essentials |
✅ / ✅ / ✅ / Pending |
| Sentinel 🛡️ |
openclaw-security-audit, afrexai-incident-response |
✅ All installed 2026-02-14 |
| Riko ⚠️ |
risk-management-specialist, afrexai-business-continuity |
✅ All installed 2026-02-14 |
Skill Installation Process
Per Skill Installation Best Practices:
- Search ClawHub for skill (downloads, stars, version)
- Read all skill files (SKILL.md, scripts, _meta.json)
- Vet using Paul/Sentinel's skill-vetter protocol
- Classify risk (LOW/MEDIUM/HIGH/EXTREME)
- Approve (HIGH+ requires Simon)
- Install to correct agent workspace
- Configure in openclaw.json
- Record in this training log
- Test with a simple task
- Re-audit quarterly
Training Records
| Date |
Event |
Skills |
Agent(s) |
Auditor |
Result |
| 2026-02-14 |
Initial skill audit |
15 skills across swarm |
All existing |
Maya (sub-agent) |
All passed — 10 LOW, 4 MEDIUM, 0 HIGH |
| 2026-02-14 |
Paul security skills installed |
skill-vetter, openclaw-security-audit, zero-trust |
Paul |
Maya |
✅ All LOW risk |
| 2026-02-14 |
Phase 2 agent creation |
Workspace files |
Laura, Kai, Iris, Achida |
— |
Workspaces created |
| 2026-02-14 |
Phase 2b skills install |
proactive-tasks, todo-management, git, afrexai-compliance-audit, milkee, finance |
Kai, Iris, Achida |
Maya (overnight) |
✅ All installed |
| 2026-02-14 |
Phase 3 agents created |
Workspace files + seo, seo-content-writer, content-writing-thought-leadership, social-media-agent, social-media-scheduler, email-marketing |
Mia, Zara |
Maya (overnight) |
✅ Created + skills installed |
| 2026-02-14 |
Phase 4 agents created |
Workspace files + openclaw-security-audit, afrexai-incident-response, risk-management-specialist, afrexai-business-continuity, gdpr-dsgvo-expert, contract, legal |
Sentinel, Riko, Felix |
Maya (overnight) |
✅ Created + skills installed |
| 2026-02-14 |
Phase 5 agents created |
Workspace files + personal-assistant, onboarding, chro, milkee, finance, customer-support, afrexai-onboarding |
Luna, Hana, Finn, Susan, Nora, Bob |
Maya (overnight) |
✅ Created + skills installed |
Review Schedule
- Quarterly: Re-audit all installed skills (Paul/Sentinel)
- On install: Every new skill vetted before deployment
- On update: Re-vet after any skill version change
- Annually: Full competence review of all agents (Maya + Simon)
References