Skip to main content

Digital Talent Bench

Talent Bench — AI Agents mapped to 2026–2030 market roles

Roles · Feedback · Teamwork · 7 Skills for Production AI Agents

44 Agents·7 Roles·ReACT·15 Skill Families·7 Production Skills·CxO Insights

Agent Role Architecture (7 Roles)

Enterprise agentic architecture — click any role to explore agents, commands & CxO value

one-way flowbidirectional
👁SUPERVISOR3 agents
💬USER INPUTHITL
📝PLANNER4 agents
⚙️DOER16 agents
📊PRESENTER2 agents
📚LEARNER4 agents
🔧TOOL OPERATOR7 agents
✅FEEDBACK4 agents

ReACT Pattern — /adlc Command Flow

Reason → Act → Observe → Answer — PDCA convergence until ≥99.5% & ≥95% agreement

📝
REASON
planner
PDCA: PlanPO + CA Coordination
Break down the problem, define requirements, design architecture
🔧
ACTION
doer
PDCA: DoSpecialist Execution
Execute implementation with tools, APIs, and code changes
✅
OBSERVE
feedback
PDCA: CheckPDCA Cross-Validation
4-way validation: MCP + Native API + Runbooks CLI + Visual
📊
ANSWER
presenter
PDCA: ActHITL Deliverable
Format results for CxO stakeholders, publish to Confluence
ReACT
PDCA Convergence Target
Autonomous loops until ≥99.5% pass rate · ≥95% agent agreement
99.5%
Pass Rate
95%
Agreement

Gap Analysis Grid

Agent-to-role coverage by category

CategoryMarket RoleADLC AgentLayerModelProficiencyDemandStatus
ArchitectureCloud Architectcloud-architectdecisionopusexpert→experthighcovered
Solutions Architectsolutions-architectdecisionopusnovice→competenthighgap_filled
OtherLifecycle Stewardlifecycle-stewarddecisionsonnetnone→competentmoderategap_filled
Digital Strategy Leaddigital-strategy-leaddigital-transformationhaikucompetent→expertmoderatecovered
Sustainability Engineersustainability-engineerdigital-transformationsonnetcompetent→expertmoderatecovered
Backstage Engineerbackstage-engineerexecutionsonnetnone→competentmoderategap_filled
Career Advisorcareer-advisorexecutionsonnetnone→competentmoderategap_filled
Commerce Engineercommerce-engineerexecutionsonnetnone→competentmoderategap_filled
Commerce FinOps Engineercommerce-finops-engineeroperationshaikunone→competentmoderategap_filled
Technical Writertechnical-writeroperationshaikucompetent→expertmoderatecovered
Project Analystproject-analystorchestratorshaikucompetent→competentmoderatecovered
Datahub Governance Engineerdatahub-governance-engineerxopssonnetnone→competentmoderategap_filled
Product ManagementProduct Ownerproduct-ownerdecisionopusexpert→experthighcovered
Market Intelligence Analystcompetitive-intelligence-analystdigital-transformationhaikucompetent→expertmoderatecovered
Business Analystbusiness-analystorchestratorshaikunovice→competenthighgap_filled
SecuritySecurity Engineersecurity-compliance-engineerdecisionopusexpert→expertcriticalcovered
Other/LeadershipDigital Transformation Leadstrategic-transformation-leaddigital-transformationsonnetcompetent→experthighcovered
Tech Leadtech-leadorchestratorssonnetcompetent→experthighcovered
SoftwareCode Reviewercode-reviewerexecutionhaikucompetent→experthighgap_filled
Developer Experience Engineerdeveloper-experience-engineerexecutionsonnetexpert→expertmoderatecovered
Full-Stack Engineerfullstack-engineerexecutionsonnetcompetent→expertcriticalcovered
Integration Engineermulesoft-engineerexecutionsonnetnone→competenthighgap_filled
Python Engineerpython-engineerexecutionsonnetexpert→experthighcovered
Salesforce Developersalesforce-engineerexecutionsonnetnone→competentcriticalgap_filled
Data & AnalyticsData Scientistdata-scientistexecutionsonnetnovice→competentcriticalgap_filled
ML Engineerml-engineerexecutionsonnetnovice→competentcriticalgap_filled
BI/Analytics Engineertableau-engineerexecutionsonnetnone→competenthighgap_filled
AI Systems Architectai-systems-architectxopssonnetcompetent→expertcriticalcovered
Data Engineerdata-engineerxopssonnetcompetent→expertcriticalcovered
Data Infrastructure Engineervector-database-engineerxopssonnetcompetent→experthighcovered
Infrastructure/SecurityDevSecOps Engineerdevops-security-engineerexecutionsonnetexpert→experthighcovered
InfrastructureInfrastructure Engineerinfrastructure-engineerexecutionsonnetexpert→experthighcovered
Platform Engineerplatform-engineerexecutionsonnetnovice→competentcriticalgap_filled
Infrastructure/CloudKubernetes Engineerkubernetes-engineerexecutionsonnetexpert→experthighcovered
Infrastructure/OpsObservability Engineerobservability-engineerexecutionsonnetexpert→experthighcovered
Software/QAQA Engineerqa-engineerexecutionsonnetexpert→experthighcovered
QA Automation Engineerqa-automation-engineeroperationshaikucompetent→experthighcovered
DesignUX Designerux-designerexecutionsonnetnone→competenthighgap_filled
Analytics & ReportingFinOps Engineerfinops-engineeroperationshaikuexpert→experthighcovered
Infrastructure/SRESite Reliability Engineersre-engineeroperationshaikucompetent→expertcriticalcovered
AgileAgile Coachagile-coachorchestratorshaikunone→competenthighgap_filled
Delivery Managerdelivery-managerorchestratorshaikucompetent→expertmoderatecovered
Scrum Masterscrum-masterorchestratorshaikunone→competentcriticalgap_filled
Software/AIPrompt Engineerprompt-engineerorchestratorssonnetexpert→expertcriticalcovered

7 Skills for Enterprise-Grade Production AI Agents

From “The 7 Skills You Need to Build AI Agents” — mapped to ADLC components

Design(1-3)
🏗
#1System Design02:40
▶

Build the orchestra: LLM making decisions, tools executing actions, databases storing state, multiple sub-agents coordinating

📜
#2Tool & Contract Design04:02
▶

Every tool has a contract — inputs, outputs, required fields. Vague contracts let the LLM fill gaps with imagination.

🔍
#3Retrieval Engineering05:15
▶

Quality of retrieval determines the ceiling of agent performance — document chunking, embedding models, and re-ranking strategy.

Reliability(4-5)
🛡
#4Reliability Engineering06:55
▶

Handle failures gracefully: retry logic, timeouts, fallback paths, circuit breakers. Agents make API calls — APIs fail.

🔒
#5Security & Safety08:24
▶

Input validation, output filters, permission boundaries. Good hygiene: does the agent really need write access?

Operations(6-7)
📈
#6Evaluation & Observability09:59
▶

You cannot improve what you cannot measure. Tracing, evaluation pipelines, success rates, latency, cost per task.

🎯
#7Product Thinking11:10
▶

Design for humans using the agent. Show confidence vs uncertainty. Decide when to ask for clarification vs escalate.

Skills Matrix

38 agents × 17 skill families — hover row or column to highlight

Agentawsazuredevelopmentfinopsgovernancearchitecturequalitysecurityobservabilityreportingvalidationtestingcdkterraformdevcontainerproductdocumentation
agile-coach
business-analyst
delivery-manager
project-analyst
prompt-engineer
scrum-master
tech-lead
cloud-architect
product-owner
security-compliance-engineer
solutions-architect
data-scientist
developer-experience-engineer
devops-security-engineer
fullstack-engineer
infrastructure-engineer
kubernetes-engineer
ml-engineer
mulesoft-engineer
observability-engineer
platform-engineer
python-engineer
qa-engineer
salesforce-engineer
tableau-engineer
ux-designer
code-reviewer
finops-engineer
qa-automation-engineer
sre-engineer
technical-writer
competitive-intelligence-analyst
digital-strategy-lead
strategic-transformation-lead
sustainability-engineer
ai-systems-architect
data-engineer
vector-database-engineer
decision
execution
operations
orchestrators
digital-transformation
xops
has skill (dim = default, bright = on hover)

Agent Role Matrix

38 agents × 7 role archetypes — filled = primary, dim = secondary

Agent👁 Supervisor📝 Planner⚙️ Doer🔧 Tool Operator📚 Learner✅ Feedback📊 PresenterModelLayerSkills Coverage
code-reviewer
product-owneropusdecisionsystem-design, product-thinking
cloud-architectopusdecisionsystem-design, tool-contract-design, reliability-engineering
security-compliance-engineeropusdecisionsecurity-safety, evaluation-observability
solutions-architectopusdecisionsystem-design, reliability-engineering
python-engineersonnetexecutiontool-contract-design, reliability-engineering
infrastructure-engineersonnetexecutionsystem-design, reliability-engineering
kubernetes-engineersonnetexecutionsystem-design, reliability-engineering
fullstack-engineersonnetexecutionproduct-thinking, tool-contract-design
devops-security-engineersonnetexecutionsecurity-safety, reliability-engineering
platform-engineersonnetexecutionsystem-design, reliability-engineering
ml-engineersonnetexecutionsystem-design, retrieval-engineering
data-scientistsonnetexecutionretrieval-engineering, evaluation-observability
salesforce-engineersonnetexecutiontool-contract-design, product-thinking
mulesoft-engineersonnetexecutionsystem-design, tool-contract-design
tableau-engineersonnetexecutionevaluation-observability, product-thinking
ux-designersonnetexecutionproduct-thinking, evaluation-observability
developer-experience-engineersonnetexecutionsystem-design, tool-contract-design
observability-engineersonnetexecutionevaluation-observability, reliability-engineering
qa-engineersonnetexecutionevaluation-observability, security-safety
scrum-mastersonnetorchestratorsproduct-thinking, evaluation-observability
delivery-managersonnetorchestratorssystem-design, product-thinking
tech-leadsonnetorchestratorssystem-design, reliability-engineering
project-analystsonnetorchestratorsretrieval-engineering, evaluation-observability
business-analystsonnetorchestratorsproduct-thinking, retrieval-engineering
prompt-engineeropusorchestratorsretrieval-engineering, product-thinking
agile-coachsonnetorchestratorsproduct-thinking, evaluation-observability
sre-engineerhaikuoperationsreliability-engineering, evaluation-observability
finops-engineerhaikuoperationsevaluation-observability, product-thinking
qa-automation-engineerhaikuoperationsevaluation-observability, tool-contract-design
technical-writerhaikuoperationsproduct-thinking, retrieval-engineering
strategic-transformation-leadsonnetdigital-transformationsystem-design, product-thinking
competitive-intelligence-analystsonnetdigital-transformationretrieval-engineering, product-thinking
digital-strategy-leadsonnetdigital-transformationsystem-design, product-thinking
sustainability-engineersonnetdigital-transformationevaluation-observability, product-thinking
ai-systems-architectsonnetxopssystem-design, retrieval-engineering
data-engineersonnetxopssystem-design, retrieval-engineering
vector-database-engineersonnetxopsretrieval-engineering, tool-contract-design
decision
execution
operations
orchestrators
digital-transformation
xops
primary (bright) / secondary (dim)

Skill Taxonomy Explorer

15 Tier-1 families · 187 skills — click a card or use the filter

Development

21 skills▶
  • Application Development
    Python, Java, Go, Rust +2
  • Web Development
    React, Next.js, Vue, Docusaurus +2
  • API Development
    FastAPI, REST, GraphQL, gRPC +1
  • CLI Development
    Click, Rich, Pydantic, argparse

Cloud

18 skills▶
  • AWS
    EC2, S3, Lambda, EKS +5
  • Azure
    Azure VMs, AKS, Azure Functions, Cost Management +1
  • Multi-Cloud
    Terraform, Pulumi, Cloud Agnostic Patterns, FOCUS 1.2+

Architecture

13 skills▶
  • Solutions Architecture
    Well-Architected Framework, Landing Zone, Microservices, Event-Driven
  • Data Architecture
    Data Modeling, ETL/ELT, Data Lake, Data Mesh +1
  • Security Architecture
    Zero Trust, IAM Design, Encryption Patterns, Threat Modeling

DevOps

12 skills▶
  • CI/CD
    GitHub Actions, ArgoCD, GitOps, Release Engineering
  • Infrastructure as Code
    Terraform, CDK, Helm, Kustomize
  • Containers & Orchestration
    Docker, Kubernetes, K3s/K3D, Service Mesh

Data Engineering

14 skills▶
  • Data Pipelines
    Spark, Airflow, dbt, Kafka +1
  • MLOps
    MLflow, Kubeflow, SageMaker, Vertex AI +1
  • Vector & RAG
    ChromaDB, Pinecone, RAG Patterns, Embeddings

Data Science

15 skills▶
  • Machine Learning
    scikit-learn, PyTorch, TensorFlow, XGBoost +1
  • NLP & LLMs
    Claude API, Prompt Engineering, Fine-tuning, RAG +1
  • Analytics
    Pandas, SQL, Visualization, Statistical Analysis +1

Security

15 skills▶
  • Compliance
    SOC2, APRA CPS 234, PCI-DSS, ISO 27001 +2
  • Application Security
    SAST, DAST, SBOM, Supply Chain +1
  • Security Operations
    Security Hub, GuardDuty, Incident Response, Forensics

Operations & Infrastructure Services

12 skills▶
  • Site Reliability
    SLOs/SLIs, Incident Management, Chaos Engineering, Runbooks
  • Observability
    MELT Telemetry, Prometheus, Grafana, CloudWatch +1
  • Platform Engineering
    Internal Developer Platforms, Developer Experience, Self-Service

Testing

9 skills▶
  • Test Automation
    pytest, Playwright, E2E Testing, BDD +1
  • Quality Engineering
    Test Strategy, Coverage Analysis, CI/CD Quality Gates, Champion-Challenger

Product Management

12 skills▶
  • Product Ownership
    User Stories, INVEST, Backlog Management, Stakeholder Management
  • Product Strategy
    OKRs, Roadmapping, Market Analysis, Competitive Intelligence
  • Business Analysis
    Requirements Engineering, Process Mapping, Data Analysis, Reporting

Agile

8 skills▶
  • Scrum & Kanban
    Sprint Planning, Retrospectives, Velocity Tracking, WIP Limits
  • Agile Coaching
    Team Facilitation, Servant Leadership, Scaled Agile, Continuous Improvement

Design

10 skills▶
  • UX Design
    User Research, Wireframing, Prototyping, Usability Testing +1
  • UI Design
    Design Systems, Figma, Responsive Design, Accessibility +1

Analytics & Reporting

10 skills▶
  • FinOps & Cost Analytics
    FOCUS 1.2+, Cost Allocation, Rightsizing, Reserved Instances +1
  • Business Intelligence
    Power BI, Vizro, Dashboards, KPI Frameworks +1

Automation

8 skills▶
  • AI Agent Systems
    Claude Code, Agent Orchestration, MCP Servers, Hooks & Rules
  • Workflow Automation
    Taskfile, Makefile, Shell Scripting, ITSM Lifecycle

Database & Data Warehouse

10 skills▶
  • Relational
    PostgreSQL, MySQL, RDS, Aurora +1
  • NoSQL & Vector
    DynamoDB, MongoDB, ChromaDB, Redis +1

Click any card to drill down into the full T2/T3 skill detail and agent mappings.

CxO Insights

How each executive persona benefits from the 37-agent talent bench

💻
CTO / VP Engineering
Executive Persona

Architecture quality, agent reliability, engineering velocity

plannersupervisordoersystem-designreliability-engineeringevaluation-observability
DORA 4/4 Elite: DF 38/30d, LT <1d, CFR 0%, MTTR 7s
Key Metric
How to Optimize

Opus for planning decisions, Sonnet for execution, golden-prompt pattern for consistency

→ See live project
💰
CFO / VP Finance
Executive Persona

Cost efficiency, FinOps governance, ROI on agent investment

tool-operatorfeedbackpresenterevaluation-observabilityproduct-thinking
37 agents replace hiring pipeline — Haiku 3x cheaper for operations tier
Key Metric
How to Optimize

Haiku for operations tasks, Sonnet for specialists, max 2h/week coordination overhead

→ See live project
🔒
CSO / CISO
Executive Persona

Security posture, compliance gates, supply chain integrity

feedbacksupervisorsecurity-safetytool-contract-design
22 hooks enforce compliance — git mutations blocked deterministically
Key Metric
How to Optimize

READONLY-by-default profiles, docker-first supply chain, SOC2/APRA CPS 234 at every gate

→ See live project
👥
HR / CPO / Talent
Executive Persona

Talent scaling, skill taxonomy, team composition optimization

learnerpresenterproduct-thinkingretrieval-engineering
15 skill families, 37 market-role-mapped agents, zero hiring lead time
Key Metric
How to Optimize

Skill taxonomy from McKinsey Rewired S2 — role_archetype standardizes team composition

→ See live project
ONE PLATFORM · FIVE SUB-PRODUCTS

ADLC Framework

ADLC v3.8.0 · 44 agents · 206 commands · 238 skills · 98 hooks · 31 rules · DORA 4/4 Elite

44 agents206 commands238 skills98 hooks31 rulesDORA 4/4 Elite
HOW IT WORKS — 7-PHASE DELIVERY LOOP
THINK
Assumptions surfaced, scope confirmed
/adlc.plan
▶
PLAN
Specs become INVEST stories
/ceremony:plan
▶
BUILD
Specialists write code with file-lock
/adlc
▶
REVIEW
Code reviewer checks quality gates
/adlc
▶
TEST
Automated + evidence-gated validation
/adlc
▶
SHIP
HITL approves · git, terraform, deploy
git commit
▶
REFLECT
DORA collected, retro evidence written
/ceremony:retro
TRIGGER TABLE — SAY IT, IT HAPPENS
You say
What happens
Who activates
"plan a sprint"
/ceremony:plan opens INVEST gate — stories scored + delegated
product-owner
"sync stories to JIRA"
/sync:jira-push creates 100+ issues with rsid: idempotency — no duplicates
atlassian-tools MCP
"run a standup"
/ceremony:standup reads DORA + blockers from live data in under 10 seconds
observability-engineer
"run a retrospective"
/ceremony:retro aggregates 4L review + DORA actuals + writes evidence
observability-engineer
"review this PR"
code-reviewer checks against 100 anti-patterns, outputs structured gaps
code-reviewer
"fix this bug"
RED-GREEN-REFACTOR cycle — failing test first, then minimal fix, then coverage
qa-engineer + python-engineer
"onboard this repo"
git submodule + symlinks + CLAUDE.md.local — every agents active in 5 minutes
developer-experience-engineer
"ship a feature"
/adlc orchestrates Think→Ship — specialists execute, HITL commits
product-owner + specialist
"check cloud spend"
READONLY AWS query executes — FOCUS-normalised cost data returned instantly
finops-engineer
"generate a cost report"
CxO-readable summary with reconciled numbers, owner + timeline — Confluence-ready
observability-engineer
SPECIALIST ROSTER — TOP 12 UNIVERSAL AGENTS
Specialist
Phase
What they do
Command
product-owner
Plan
INVEST gate, story prioritisation, business value validation
/adlc.plan
cloud-architect
Plan
Architecture decisions, ADRs, deployment strategy
/adlc.plan
python-engineer
Build
PyPI packages, CLI tools, API connectors, ETL pipelines
/adlc
fullstack-engineer
Build
React pages, Docusaurus docs, terminal UI, Rich CLI
/adlc
infrastructure-engineer
Build
Terraform modules, CI/CD pipelines, IaC best practices
/adlc
data-engineer
Build
Snowflake DDL, dbt pipelines, event streaming, bulk load
/adlc
security-compliance-engineer
Review
APRA CPS 234/230, SAST, audit trail, cert inventory
/adlc
code-reviewer
Review
100 anti-pattern checks, coding discipline, PR quality gates
/adlc
qa-engineer
Test
RED-GREEN-REFACTOR, NFR baselines, 4-way cross-validation
/adlc
sre-engineer
Test
DORA metrics, MELT telemetry, alert thresholds, hypercare
/adlc
observability-engineer
Reflect
Ceremonies, DORA collection, sprint evidence, CxO reports
/ceremony:retro
finops-engineer
Reflect
Cloud cost optimisation, FOCUS transforms, anomaly detection
/adlc
ENGINEERING PHILOSOPHY
Test-first (RED-GREEN-REFACTOR)
Write a failing test before any implementation code. Rule 6 in coding-discipline.md — prevents testing theater where 128/128 tests pass but the wheel crashes.
Systematic over ad-hoc (PDCA)
Define success criteria before first edit, loop until verified. Rule 4 — transforms 'make it work' into a verifiable goal with evidence.
Evidence over claims (no NATO)
detect-nato-violation.sh blocks any 'complete/done/finished' without a path in tmp/. Completion is proved, not reported.
Simplicity first (KISS + surgical edits)
Maximum 200 LOC per module, no abstractions for single-use code (Rule 2), every changed line traces to the request (Rule 3).
ADOPT IN 3 STEPS — under 5 minutes
Add the framework as a git submodule. All 44 agents, 206 commands, 31 rules, and 98 hooks become active immediately — scoped to your project via CLAUDE.md.local.
1
Add submodule
git submodule add [email protected]:1xOps/adlc-framework.git .adlc
Pin to a tag — e.g. v3.8.0
2
Symlink config
ln -s .adlc/.claude .claude && ln -s .adlc/.specify .specify
Agents and rules become active immediately
3
Override personas
cp .adlc/.claude/templates/CLAUDE.md.local.template CLAUDE.md.local
5-min edit: personas, AWS profiles, quality gates, PII boundary
ADOPT vs BUILD — what's free, what you own
WHAT YOU GET FREE (framework)
WHAT YOU BUILD (your product)
AGENT SPECIALISTS
12+ pre-trained agents (data-engineer, ml-engineer, security-compliance-engineer, qa-engineer, sre-engineer, product-owner)
 
Domain personas — Snowflake architect, Energy SME, Smart Meter specialist, B2B API owner
SPRINT AUTOMATION
Ceremony, planning and sync commands (/adlc.plan, /ceremony:*, /sync:jira-push, /metrics:update-dora)
 
Product-prefix stories, domain Confluence space, domain-specific Gantt CSV
ENFORCEMENT HOOKS
7 critical hooks (enforce-coordination, detect-nato, testing-theater, specialist-delegation, component-counts)
 
Domain-specific quality gates (dbt test gates, SQL schema validation, pipeline quality checks)
GOVERNANCE RULES
Governance rules — authority chain, anti-patterns catalogue, coding discipline, agency boundaries
 
Domain compliance rules (smart meter privacy, NERC CIP-013, regulatory audit trail)
ATLASSIAN INTEGRATION
JIRA/Confluence sync — 39 templates, 101-story CSV pattern, 2-board model (SPM + OPS)
 
Multi-tenant data isolation, customer billing reconciliation, API rate limit governance
INFRASTRUCTURE PATTERNS
Terraform AWS/Azure, FinOps cost gates, security audit automation, FinOps FOCUS schema
 
Snowflake schemas, dbt project structure, lineage assertions, event streaming topology
EVIDENCE STRUCTURE
tmp/ layout, coordination logs, test result schemas, ceremony artifacts
 
Domain ceremonies — data quality check reports, dbt lineage reviews, parallel-run evidence
VERSION TESTED ON
v3.8.0 — tested on AIOps (RAG chatbot), CloudOps (50+ AWS accounts), FinOps · FOCUS (multi-cloud cost)
 
Your product stack — B2B SaaS, multi-customer isolation, domain-specific event patterns
ENGINEERING DISCIPLINE · KARPATHY RULES · ANTI-PATTERNS · GOVERNANCE

Engineering Discipline

Every agent on the talent bench applies these engineering disciplines. Source references link to the canonical .claude/ SSOT — the content lives there; this page links, not re-hosts.

38-Agent Talent Bench · 6 Tiers
Source: CLAUDE.md Agent Team (7 Roles · ReACT) · Full gap analysis and agent matrix: Gap Analysis Grid ↑ ·Agent Role Matrix ↑
Decision Layer
opus
4
product-ownercloud-architectsecurity-compliance-engineersolutions-architect
Execution Layer
sonnet + 1 haiku
17
developer-experience-engineerpython-engineerinfrastructure-engineerkubernetes-engineerqa-engineerobservability-engineerfullstack-engineerdevops-security-engineerplatform-engineerml-engineerdata-scientistux-designerdata-engineersalesforce-engineermulesoft-engineertableau-engineercode-reviewer
Orchestrators
sonnet/haiku + 1 opus
7
prompt-engineertech-leadproject-analystdelivery-managerbusiness-analystagile-coachscrum-master
Operations Tier
haiku
4
sre-engineerfinops-engineerqa-automation-engineertechnical-writer
Digital Transformation
sonnet + haiku
4
strategic-transformation-leadsustainability-engineercompetitive-intelligence-analystdigital-strategy-lead
XOps
sonnet
2
vector-database-engineerai-systems-architect
7 Karpathy Coding Discipline Rules
Canonical SSOT: .claude/rules/engineering/coding-discipline.md · Every agent applies all 7 rules.
R1BLOCKS
Think Before Coding
Surface assumptions; ask don't guess; present alternatives before implementing.
HIDDEN_ASSUMPTION
R2BLOCKS
Simplicity First
Min code to solve stated problem. No Strategy/Factory for 1 implementation.
PREMATURE_ABSTRACTION
R3BLOCKS
Surgical Changes
Touch only what was requested. Every changed line traces to the request.
DRIVE_BY_REFACTORING
R4BLOCKS
Goal-Driven PDCA
Define success criteria in testable form before writing any code.
SUCCESS_CRITERIA_MISSING
R5BLOCKS
Subagent Context Preservation
Command files = sole delivery channel to subagent execution context. Never strip.
DOMAIN_CONTENT_STRIPPING
R6BLOCKS
RED-GREEN-REFACTOR
Write failing test FIRST. Min code to pass. Refactor under green only.
TESTING_THEATER
R7BLOCKS
Try Autonomous First
Attempt autonomous execution before deferring. Hook denial = only valid stop.
LAZY_DEFERRAL
Anti-Patterns Top-20 (of 100+)
Canonical SSOT: .claude/rules/governance/anti-patterns-catalog.md · Full catalog: 100+ patterns with hook enforcement.
PATTERNSEVERITYDESCRIPTION
STANDALONE_EXECUTIONBLOCKINGSpecialist work without PO+CA approval
NATO_VIOLATIONBLOCKINGCompletion claims without evidence paths in tmp/
COORDINATION_WITHOUT_DELEGATIONBLOCKINGPO+CA ran but raw Edit/Write used instead of specialist Task
SCOPE_DRIFT_BYPASSRULES-LAYERPrior-scope coordination logs reused for new scope
TEXT_OUTPUT_BYPASSRULES-LAYERImplementation content in text without PO+CA coordination
FALSE_AS_IS_STATERULES-LAYERPlanning on fabricated baseline not cross-validated
KISS_MISINTERPRETATIONRULES-LAYERKISS applied to evidence richness instead of code complexity
NARROW_SEARCH_SCOPERULES-LAYERPer-account search when org-wide discovery tools exist
JSON_ONLY_NATORULES-LAYEREvidence JSON updated without updating HITL-consumable deliverable
AGENT_GIT_COMMIT_SUGGESTIONBLOCKINGAgent suggests or attempts git mutations — violates Principle I
TESTING_THEATERBLOCKINGHigh pass counts but mocks without assertions, inflated omit
RACE_CONDITION_SCORINGRULES-LAYERConcurrent scoring agents see different intermediate states
SCORING_THEATERBLOCKINGRe-scoring unchanged artifacts — PDCA count increments, nothing changes
NO_ESTIMATED_COUNTSRULES-LAYERNumeric claims by estimation, not verification command output
AWS_PROFILE_SEMANTIC_MISMATCHRULES-LAYERWrong profile for API — each profile has exactly ONE correct purpose
MOCK_CIRCULAR_VALIDATIONBLOCKINGTest asserts hardcoded mock value that code also returns — circular
PARALLEL_AGENT_SCOPE_BLEEDRULES-LAYERParallel agents create/modify files outside assigned scope
PLAN_DELIVERY_MISMATCHRULES-LAYERSprint plan claims features not in shipped code
LAZY_DEFERRALRULES-LAYERClaiming deferred without attempting autonomous execution first
HIDDEN_ASSUMPTIONRULES-LAYERAgent picks interpretation silently without surfacing assumption
Golden-Prompt Pattern (9-Part Anatomy)
Every specialist delegation uses the 9-part positive-framing structure. Prevents FRAGMENTED_PROMPT_INJECTION and NEGATIVE_FRAMING_FIXATION.
.claude/skills/governance/golden-prompt-pattern/SKILL.md
System-Thinking Governance Lens
Applied before any architectural or governance recommendation. Pairs with the Critical Thinking Lens practitioner checklist for dual-lens evaluation.
.claude/skills/governance/system-thinking/SKILL.md
Practitioner checklist: Critical Thinking Lens →

Browse the Agent Marketplace

All 37 enterprise agents available as installable components

Explore in Marketplace →