peterbamuhigire Capability Matrix
Acknowledgement: Shared by Peter Bamuhigire, techguypeter.com, +256 784 464178.
Use When
- Starting a new project and choosing the initial skill stack.
- Scoping a feature in an existing codebase and deciding which skills should load alongside the platform skill.
- Reviewing a PR or design and asking "what should have been loaded for this work?".
- Onboarding to a new domain (mobile, AI, K8s, GIS) and wanting the canonical recommended set rather than discovering skills one by one.
Do Not Use When
- The work is purely conceptual (writing prose, planning, brainstorming) and no implementation is imminent. Skills are loaded for execution; the matrix is a tool for execution work.
- The domain is genuinely novel (no row applies). Flag the gap and ask the user to clarify; do not guess by stretching an unrelated row.
- A specialist skill's own
Use Whenalready covers the situation completely. The matrix is for stack composition, not single-skill triggering.
Routing Intake Guidance
- The technology domain or domains in scope (web, mobile, API, AI, etc.).
- A clear sense of whether the work is greenfield or brownfield (changes which Foundation skills apply).
- Awareness that Validation entries here are pulled from
validation-contract's seven evidence categories.
Routing Method Detail
- Identify the matrix row(s) that match the work.
- Load every skill in the row's Foundation column first, then Implementation, then Validation.
- Add Companions as the work demands per
references/companion-rules.md. - If two rows could apply (e.g., AI feature inside a multi-tenant SaaS), load both rows and dedupe.
Quality Standards
- Treat the matrix as the minimum starting set, not the cap. Add more skills if the work warrants.
- Never load a skill from the matrix without also reading its
Use Whento confirm the situation actually matches. - Flag missing rows. The matrix is not exhaustive; an unknown domain is a signal to ask, not to improvise.
Legacy Routing Pitfalls
- Loading every skill in a row without reading their
Use Whentriggers — the matrix recommends, the skill triggers gate. - Treating Validation column entries as the only validation needed —
validation-contractdefines the full evidence requirement; the matrix lists the canonical contributors per domain. - Inventing a new row inline rather than updating this skill. The matrix is the single source of truth.
Routing Deliverable Detail
- A loaded skill stack tailored to the domain, ordered Foundation → Implementation → Validation → Companions.
- A clear statement when the matrix has no row for the work, prompting the user to clarify the domain.
References
- references/domain-rationale.md: per-row rationale for the skill picks.
- references/companion-rules.md: when each Companion category kicks in (UX, content, security baseline, release).
How to use the matrix
- Identify the domain. Ask: "what is this project / feature primarily?" Match it to one row. If two rows apply (a common case — e.g., "AI feature in a SaaS"), load both rows and dedupe the skill list.
- Load Foundation → Implementation → Validation in order. Foundation skills set up the architecture, contracts, and data shape decisions before framework-level work begins. Implementation skills handle the platform-specific build. Validation skills carry the ship-readiness gates.
- Add Companions as the work demands. Companions are cross-cutting skills that often apply but not always — see
references/companion-rules.mdfor the trigger rules.
The matrix
| Domain | Foundation | Implementation | Validation | Companions |
|---|---|---|---|---|
| Web/SaaS | system-architecture-design, database-design-engineering |
php-modern-standards, nodejs-development |
vibe-security-skill, web-app-security-audit, frontend-performance |
git-collaboration-workflow, ux-writing |
| Multi-tenant SaaS | multi-tenant-saas-architecture, database-design-engineering, saas-erp-system-design |
php-modern-standards, dual-auth-rbac, modular-saas-architecture |
vibe-security-skill, observability-monitoring |
saas-business-metrics, subscription-billing, stripe-payments, feature-planning |
| iOS | ios-development, ios-architecture-advanced |
swiftui-design, ios-data-persistence, ios-networking-advanced |
ios-tdd, ios-app-security, app-store-review |
ios-pdf-export, ios-biometric-login, ios-rbac, ios-monetization |
| Android | android-development |
jetpack-compose-ui, android-data-persistence, android-room |
android-tdd, google-play-store-review |
android-pdf-export, android-biometric-login, mobile-rbac, mobile-custom-icons |
| KMP / Compose Multiplatform | kmp-development |
kmp-compose-multiplatform |
kmp-tdd |
ios-development, android-development |
| API / HTTP | api-design-first, system-architecture-design |
nodejs-development, php-modern-standards |
api-testing-verification, vibe-security-skill, graphql-security |
api-pagination, api-error-handling, realtime-systems |
| Database (relational) | database-design-engineering, mysql-data-modeling |
mysql-best-practices, postgresql-fundamentals |
database-reliability, mysql-query-performance, postgresql-performance |
mysql-administration, postgresql-administration, database-internals |
| Frontend (React / Next) | nextjs-app-router, system-architecture-design |
react-development, react-patterns, tailwind-css |
frontend-performance, design-audit |
grid-systems, practical-ui-design, interaction-design-patterns, motion-design, responsive-design, ux-writing, ai-slop-prevention |
| AI Feature | ai-app-architecture, ai-feature-spec, ai-architecture-patterns |
ai-llm-integration, ai-prompt-engineering |
ai-evaluation, llm-security, ai-error-handling, ai-security |
ai-ux-patterns, ux-for-ai, ai-cost-modeling, ai-slop-prevention |
| LLM Integration | ai-llm-integration, ai-app-architecture |
claude-api, deepseek-integration, openai-agents-sdk |
llm-security, ai-evaluation, ai-error-handling |
ai-prompt-engineering, ai-rag-patterns, ai-agents-tools, microservices-ai-integration |
| Python Service | python-modern-standards |
python-saas-integration |
advanced-testing-strategy |
python-data-analytics, python-document-generation, python-ml-predictive, python-data-pipelines |
| TypeScript Stack | typescript-mastery, typescript-effective |
typescript-full-stack, typescript-design-patterns |
api-testing-verification, advanced-testing-strategy |
nodejs-development, react-development |
| Kubernetes | kubernetes-fundamentals, system-architecture-design |
kubernetes-production |
observability-monitoring, reliability-engineering, network-security |
kubernetes-saas-delivery, cicd-devsecops, cloud-architecture |
| GIS | gis-mapping, gis-postgis-backend |
gis-maps-integration |
vibe-security-skill, frontend-performance |
gis-enterprise-domain, database-design-engineering |
| ERP / Business System | saas-erp-system-design, multi-tenant-saas-architecture |
modular-saas-architecture, inventory-management, saas-accounting-system |
vibe-security-skill, observability-monitoring, advanced-testing-strategy |
form-ux-design, manual-guide, sdlc-design, feature-planning |
| CI/CD pipeline | cicd-pipeline-design |
cicd-pipelines, cicd-jenkins-debian |
cicd-devsecops, deployment-release-engineering |
cloud-architecture, kubernetes-fundamentals |
| Observability / SRE | observability-monitoring |
reliability-engineering, distributed-systems-patterns |
advanced-testing-strategy, deployment-release-engineering |
linux-security-hardening, network-security, database-reliability |
Domain disambiguation
When two rows could apply, load both and dedupe. Common combinations:
- AI feature inside a SaaS → AI Feature row + Multi-tenant SaaS row.
- Mobile app talking to a backend → iOS or Android row + API/HTTP row.
- GIS dashboard inside a web app → GIS row + Frontend (React/Next) row.
- AI-powered analytics in a SaaS → AI Feature row + Multi-tenant SaaS row + (if Python-based) Python Service row.
- K8s-deployed multi-tenant SaaS → Multi-tenant SaaS row + Kubernetes row + Observability/SRE row.
- Brownfield refactor of any kind → relevant domain row + always add
git-collaboration-workflowandsystem-architecture-design.
Vertical addenda
Some verticals add a small set of skills on top of whatever rows apply:
- Healthcare → add
healthcare-ui-design. - POS / Restaurant → add
pos-sales-ui-design,pos-restaurant-ui-standard. - Payments → add
stripe-payments,subscription-billing. - Auth surface → add
dual-auth-rbac,mobile-rbac,ios-rbacas platform demands. - East-African / Uganda compliance → add
uganda-dppa-compliance,dpia-generator,language-standards,east-african-english. - Reporting / dashboards → add
data-visualization,mobile-reports,mobile-report-tables,report-print-pdfas platform demands. - Photo / image handling → add
photo-management,image-compression. - Real-time features → add
realtime-systems. - PDF / document generation → add
report-print-pdf,python-document-generation,ios-pdf-export,android-pdf-export,professional-word-output,excel-spreadsheetsas needed. - Spec / SDLC docs → add
spec-architect,sdlc-design,sdlc-planning,sdlc-testing,sdlc-user-deploy,sdlc-maintenance,sdlc-post-deployment. - Systems / process / requirements description → add
systems-process-requirementswhen work needs system boundaries, workflows, scope, interfaces, data architecture, business rules, states, traceability, or design-system documentation.
Evidence-contract note
The Validation column in every row pulls skills that are themselves declared validation-contract contributors. This matrix tells you which validation skills apply per domain; validation-contract defines what evidence those skills must produce at ship time. Always load validation-contract alongside the matrix when shipping is in scope.
Companion Skills
Workflow
- Discover active entrypoints and identify the task's domain, phase, and deliverable.
- Select the minimum foundation, implementation, and validation route.
- Stop when no active skill matches or required capabilities are unavailable.
- Recover with a qualified route and named gap instead of inventing an entrypoint.
- Add companions only for mandatory cross-domain controls.
Outputs
| Artefact | Consumer | Acceptance condition |
|---|---|---|
| Ordered minimum skill route | Orchestrator and task owner | Every selected skill has a stated role and no neighbour is loaded redundantly |
Evidence Produced
| Evidence | Consumer | Acceptance condition |
|---|---|---|
| Discovery record and routing rationale | Maintainer and reviewer | All route names resolve to active entrypoints and every companion has a documented need |
Capability Contract
Minimum capability is read-only access to active skill roots and task context. Changing routers, skills, files, or permissions requires explicit authorisation.
Routing Correction Examples
- Loading every related skill. Fix: choose the minimum complete stack.
- Trusting a cached catalogue. Fix: discover active files at routing time.
- Routing by keyword alone. Fix: use intent, phase, and output.
- Inventing an entrypoint. Fix: report a capability gap.
- Omitting validation. Fix: include the applicable release gate.
Routing Scenario
A dataset-cleaning request routes to the data-quality pipeline and validation gate, not the entire report-production and orchestration stack.
Companion Skills
Inputs
| Input | Source/provider | If absent |
|---|---|---|
| Task, domain, lifecycle phase, deliverable | Requester and repository | Ask for the missing domain or route provisionally |
| Available skills and capabilities | Filesystem discovery and runtime | Report unmatched capability; do not invent a skill |
Degraded Mode
Fallback when catalogue or runtime state is incomplete: return the smallest qualified route with missing capabilities named and mark unavailable checks not assessed.
Decision Rules
| Choice | Action | Failure/risk avoided |
|---|---|---|
| One domain clearly owns the task | Route foundation to validation | Skill sprawl |
| Cross-domain control is mandatory | Add one companion | Missing gate |
| No matching active skill exists | Report a capability gap | Fabricated route |
Anti-Patterns
- Loading every related skill. Fix: choose the minimum complete stack.
- Trusting a cached catalogue. Fix: discover active files at routing time.
- Routing by keyword alone. Fix: use intent, lifecycle phase, and output.
- Inventing an entrypoint. Fix: report the capability gap.
- Omitting validation. Fix: include the applicable release gate.
Worked Example
A dataset-cleaning request routes to the data-quality pipeline and its validation gate, while unrelated report-production and research-orchestration skills remain unloaded.
Companion Skills
world-class-engineering— the production-readiness bar that the matrix's recommendations operate within.validation-contract— defines the seven evidence categories that the Validation column draws from.skill-composition-standards— house style and Inputs/Outputs contracts; load when authoring new skills that should be added to the matrix.