New Capability: AI Team Job Description Library
Date: March 9, 2026 (consolidating work from Mar 8-9) Origin: Chris requested comprehensive job descriptions after the AI org grew to 35+ agents. The org chart (built in the same session) showed the structure, but each role needed detailed documentation covering both the human-facing "when do I use this?" and the AI-facing "how does this technically work?" perspectives. Impact: Every role in the AI organization now has a dual-audience job description — 44 markdown files for AI consumption and 44 HTML files for human orientation. Contributors can answer "what does this agent do?" and "when would I use it?" by opening a styled HTML page. Agents can load their own JD for context.
What Was Built
A comprehensive job description library covering every role in the AI organization: 3 AI leaders, 1 Advisory Committee page, 36 active agents, 2 specced agents, and 1 gap roles tracker. Each role has both a markdown JD (for AI agents to load as context) and a styled HTML JD (for humans to read in a browser). That is 88 total files.
The HTML JDs follow a consistent template designed for two audiences in a single document. The top half is "For Humans" — when to engage this agent, what you will get, how it works, autonomy level, and Key Value Indicators with VP dimension mapping. The bottom half is "For AI" — activation triggers, skills loaded, upstream/downstream agents, technical processing steps, and dependencies. This dual-audience structure means a contributor reads the human section to understand when they would invoke an agent, while an AI agent reads the technical section to understand its own processing pipeline.
A deliberate naming convention distinguishes the two formats. Markdown files use functional names matching the monorepo directory structure (transcript-processor.md, session-brief.md, relationship-sentinel.md). HTML files use persona names matching the org chart (scribe.html, herald.html, sentinel.html). This means AI agents find their JD by functional name, and humans navigate by the persona names they see on the org chart. Both link back to the interactive org chart, and the org chart links forward to every JD.
Infrastructure Changes
| Change | Before | After |
|---|---|---|
| Agent documentation | AGENT.md files in monorepo only | AGENT.md + markdown JD + styled HTML JD per role |
| Human orientation | Read AGENT.md (technical, no styling) | Styled HTML page with "For Humans" section, identity card, leadership commentary |
| KVI visibility | Scattered across individual AGENT.md files | Consolidated per role with VP dimension mapping and explicit anti-patterns |
| Cross-agent connections | Described in prose within AGENT.md | Structured "receives from / feeds to" connections tables |
| Leadership perspective | Not documented anywhere | V, Sage, and Pax commentary cards on every role |
| Org chart integration | Standalone HTML page | Bidirectional: org chart nodes link to JDs, JD headers link back to org chart |
| Gap tracking | Mentioned in roadmap discussions | Dedicated gap roles page with status, build history, and remaining gaps |
| Agent identity | Name and AGENT.md only | Identity card with philosophy, origin story, and self-description |
Implementation
| File | Purpose |
|---|---|
/mnt/d/Leadership/assessments/job-descriptions/*.md (44 files) |
Markdown JDs for AI consumption |
/mnt/d/Leadership/assessments/job-descriptions/*.html (44 files) |
Styled HTML JDs for human orientation |
/mnt/d/Leadership/assessments/2026-03-08-ai-org-chart.html |
Org chart with jdMap linking to all JD pages |
Coverage
AI Leaders (3)
| Role | File (md) | File (html) |
|---|---|---|
| V (COO) | v-coo.md |
v-coo.html |
| Sage (CCO) | sage-cco.md |
sage-cco.html |
| Pax (CFO) | pax-cfo.md |
pax-cfo.html |
Advisory Committee (1)
| Role | File (md) | File (html) |
|---|---|---|
| Advisory Committee | advisory-committee.md |
advisory-committee.html |
Client Delivery — V (10)
| Agent | Persona | File (md) | File (html) |
|---|---|---|---|
| Transcript Processor | Scribe | transcript-processor.md |
scribe.html |
| Status Reporter | Herald | status-reporter.md |
herald.html |
| Document Creator | Forge | document-creator.md |
document-creator.html |
| Context Sync | Sync | context-sync.md |
context-sync.html |
| Session Brief | Beacon | session-brief.md |
herald.html |
| HubSpot Implementation PM | Architect | hubspot-implementation-pm.md |
hubspot-implementation-pm.html |
| Client Onboarding | Settler | client-onboarding.md |
client-onboarding.html |
| Client Comms | Correspondent | client-comms.md |
client-comms.html |
| Project Management | Navigator | project-management-agent.md |
marshal.html |
| Portal Manager | Pavilion-PM | portal-manager.md |
portal-manager.html |
Content & Media — V (10)
| Agent | File (md) | File (html) |
|---|---|---|
| Content Multiplier | content-multiplier.md |
content-multiplier.html |
| Content Vault | content-vault.md |
content-vault.html |
| Content Index | content-index.md |
content-index.html |
| Image Generator | image-generator.md |
image-generator.html |
| Media Management | media-management.md |
media-management.html |
| Media Pre-Production | media-pre-prod.md |
media-pre-prod.html |
| Media Post-Production | media-post-prod.md |
media-post-prod.html |
| Media Distribution | media-distribution.md |
media-distribution.html |
| Alert Manager | alert-manager.md |
alert-manager.html |
| Feedback Loop | feedback-loop.md |
feedback-loop.html |
Platform — V (7)
| Agent | File (md) | File (html) |
|---|---|---|
| Portal Manager | portal-manager.md |
portal-manager.html |
| Website Health | website-health.md |
website-health.html |
| CTA Optimization | cta-optimization.md |
cta-optimization.html |
| Code Butler | code-butler.md |
code-butler.html |
| Doc Butler | doc-butler.md |
doc-butler.html |
| Background Workers | background-workers.md |
background-workers.html |
| Self-Healing | self-healing.md |
self-healing.html |
Self-Improvement — V (5)
| Agent | File (md) | File (html) |
|---|---|---|
| Pattern Memory | pattern-memory.md |
pattern-memory.html |
| Instruction Optimizer | instruction-optimizer.md |
instruction-optimizer.html |
| Memory Librarian | memory-librarian.md |
memory-librarian.html |
| Data Integrity Checker | data-integrity-checker.md |
audit.html |
| Skill Eval | — | — |
Relationship Intelligence — Sage (6+1)
| Agent | Persona | File (md) | File (html) |
|---|---|---|---|
| Session Brief | Herald | session-brief.md |
herald.html |
| Relationship Sentinel | Sentinel | relationship-sentinel.md |
sentinel.html |
| Contributor Intelligence | Prism | contributor-intelligence.md |
contributor-intelligence.html |
| Signal Progression | Tide | signal-progression.md |
signal-progression.html |
| New Business Intelligence | Scout | new-business-intelligence.md |
new-business-intelligence.html |
| Office Hours Intelligence | Quorum | office-hours-intelligence.md |
office-hours-intelligence.html |
Financial Intelligence — Pax (2)
| Agent | Persona | File (md) | File (html) |
|---|---|---|---|
| Business Health Monitor | Pulse | business-health-monitor.md |
pulse.html |
| Renewal Tracker | Horizon | renewal-tracker.md |
renewal-tracker.html |
Cross-Functional (2)
| Agent | File (md) | File (html) |
|---|---|---|
| Convergence Analyzer | convergence-analyzer.md |
convergence-analyzer.html |
| Coach Chris | coach-chris.md |
coach-chris.html |
Gap Roles (1)
| Page | File (md) | File (html) |
|---|---|---|
| Gap Roles Tracker | gap-roles.md |
gap-roles.html |
Remaining gap: Revenue Scenario Modeler (Pax) — forward-looking financial modeling, not yet built.
HTML Template Structure
Every HTML JD contains these sections in order:
- Header — Role name, leader badge (color-coded: blue for V, teal for Sage, amber for Pax), group badge, status badge (green for Active, blue for Operational, amber for Specced)
- Identity Card — The agent's self-description, philosophy quote, and origin story
- Standup Role — What this agent contributes to daily-ops (or "Not a standup agent" for reactive roles)
- For Humans — When to engage, what you will get, how it works, autonomy level, schedule
- Key Value Indicators — KVI cards with VP dimension mapping, measurement description, and explicit anti-pattern
- For AI — Activation triggers, skills loaded, upstream/downstream agents, technical processing steps
- Current State — Honest operational assessment
- Connections Table — Structured "receives from" and "feeds to" showing data flow
- Leadership Commentary — Cards from V, Sage, and Pax with each leader's perspective on the role
- Footer — Filed date, org chart link, implementation status
Each HTML file is self-contained with embedded CSS (no external stylesheets). This means any file can be opened in a browser independently.
Usage
# Open any HTML JD in a browser
# From Windows: navigate to D:\Leadership\assessments\job-descriptions\scribe.html
# Open the org chart (links to all JDs)
# D:\Leadership\assessments\2026-03-08-ai-org-chart.html
# AI agent loading its own JD for context
# Read /mnt/d/Leadership/assessments/job-descriptions/transcript-processor.md
# Count all files
ls /mnt/d/Leadership/assessments/job-descriptions/*.md | wc -l # 44
ls /mnt/d/Leadership/assessments/job-descriptions/*.html | wc -l # 44
The org chart serves as the index. Every node links to its JD page via the jdMap object in the chart's JavaScript. Clicking an agent node opens the corresponding HTML JD.
Leader Applications
V (Operations)
V manages the largest org — 32 agents across 4 groups (Client Delivery, Content & Media, Platform, Self-Improvement). The JD library gives V a reference for every agent's triggers, KVIs, and handoff patterns. When debugging why intelligence is not flowing correctly between agents, V can check the connections table to trace the data path. When onboarding contributors, V can share the HTML pages as orientation material — the "For Humans" sections answer the question contributors always ask: "When would I use this?" The KVI cards with anti-patterns serve as enforcement: they make explicit what each agent measures and, more importantly, what would be a misuse of that measurement.
Sage (Customer)
Sage's 7 relationship intelligence agents each have detailed JDs documenting their role in the intelligence chain. The KVI mapping to VP dimensions makes explicit what each agent measures and — critically — what the anti-pattern is. This prevents relationship intelligence from drifting toward volume metrics. The connections tables show how Tide feeds into Sage's pipeline visibility, how Quorum's attendance data feeds Scout's multi-signal detection, and how Herald consumes intelligence from Sentinel and Pulse to prepare session briefs. For new contributors wondering "how does Sage know what's happening with a client?", the JD library traces the complete intelligence flow from raw signal to actionable brief.
Pax (Finance)
Pax's 2 agents (Pulse for commercial health, Horizon for renewal proximity) now have JDs that document the financial intelligence methodology. The gap roles page makes visible that Revenue Scenario Modeler is the only remaining capability gap — useful context for roadmap discussions. Pax can reference Pulse's JD when explaining what goes into the commercial health score, and Horizon's JD documents the exact renewal proximity calculation (closedate + contract length months) so the methodology is auditable.
Dependencies
| Dependency | Status | Notes |
|---|---|---|
| AGENT.md files | Confirmed | Source data for all agent JDs; 36+ AGENT.md files in monorepo |
| Org chart | Confirmed | JDs link back to org chart; org chart links to JDs via jdMap |
| CSS (self-contained) | Confirmed | Each HTML file contains its own embedded CSS; no external stylesheets |
| AI Leader identities | Confirmed | V, Sage, Pax identity prompts sourced from /mnt/d/{V,Sage,Pax}/ |
| VP dimension property names | Confirmed | All KVIs map to actual HubSpot property names from property-index |
Verification
# Count all JD files (expect 44 each)
ls /mnt/d/Leadership/assessments/job-descriptions/*.md | wc -l
ls /mnt/d/Leadership/assessments/job-descriptions/*.html | wc -l
# Verify org chart exists and links to JDs
ls /mnt/d/Leadership/assessments/2026-03-08-ai-org-chart.html
# Open any HTML JD in a browser and verify sections:
# - Header with role name + badges
# - Identity card with philosophy + origin
# - Standup Role section
# - For Humans (when to engage, what you get, how it works, autonomy)
# - KVI cards with VP dimension + anti-pattern
# - For AI (triggers, skills, upstream/downstream)
# - Current State
# - Connections table
# - Leadership Commentary (V, Sage, Pax)
# - Footer with filed date + org chart link
Design Decisions
Dual naming convention. Markdown files use functional names (transcript-processor.md) matching the monorepo agent directory. HTML files use persona names (scribe.html) matching the org chart. This serves both audiences without confusion — AI agents find their JD by the name they already know, humans navigate by the names they see on the org chart.
Self-contained CSS. Each HTML file embeds its own styles rather than referencing an external stylesheet. This means any single file can be opened, shared, or bookmarked without dependencies. The tradeoff is that a template change requires updating all 44 HTML files, but the benefit is zero deployment complexity.
Anti-patterns on every KVI. Every Key Value Indicator includes an explicit anti-pattern (e.g., Scribe's Intelligence Extraction KVI has anti-pattern "Not: transcripts processed/week"). This is enforcement built into the documentation itself. It prevents the natural drift toward volume metrics that training data patterns constantly push toward.
Leadership commentary cards. Every JD includes perspective from all three leaders, not just the agent's direct leader. This cross-org visibility ensures V knows what Sage thinks about a relationship intelligence agent, Sage knows what Pax sees in a financial agent's output, and Pax understands V's operational perspective on platform agents. The commentary is not filler — each card states what the leader specifically values or depends on from that role.
Filed: March 9, 2026 | V (COO)