cosmix Loom Skills Catalog Loader
Overview
Nine loom mechanics skills stay installed into ~/.claude/skills/, indexed by
Claude Code with their description: kept resident in every request. The
other fifty-three domain skills install instead into
~/.claude/loom-skill-catalog/<name>/SKILL.md — a directory Claude Code does
not scan, so they cost nothing resident. This skill is the bridge: it loads
one of those catalogued skills on demand.
How to Load
Invoked with an argument, read ~/.claude/loom-skill-catalog/<argument>/SKILL.md
in full with the Read tool and follow it as if it had been loaded directly.
The argument is the full skill name, for example loom-rust.
Several names may be passed at once, separated by whitespace or commas, for
example loom-ci-cd loom-rust. Split the argument and load EACH name by the
rules below, in the order given. Never treat the whole string as one skill
name — ~/.claude/loom-skill-catalog/loom-ci-cd loom-rust/SKILL.md is not a
path that can exist, and reporting it as a missing catalog is wrong.
If a name's path does not exist, try ~/.claude/skills/<name>/SKILL.md — an
install made with --skills all keeps every skill in the indexed directory
and has no catalog directory at all.
If neither path exists for a name, say which name did not resolve and continue
with the ones that did. Report the catalog as not installed ONLY when
~/.claude/loom-skill-catalog/ is itself absent AND the name is missing from
~/.claude/skills/ — a single unresolved name means a bad name, not a broken
install.
Invoked with no argument, show the table below and pick the skill that fits the task.
Catalog
| Skill | Summary |
|---|---|
loom-accessibility |
Web accessibility, WCAG compliance, inclusive design |
loom-api-design |
Designs REST/GraphQL/RPC APIs for consistency and scale |
loom-api-documentation |
OpenAPI/Swagger docs, auth flows, SDK guides |
loom-argocd |
GitOps CD for Kubernetes with Argo CD |
loom-auth |
OAuth2/JWT/RBAC auth patterns, sessions, MFA |
loom-background-jobs |
Job queues, scheduled jobs, worker pools, retries |
loom-caching |
Caching strategies: cache-aside, TTL, stampede prevention |
loom-ci-cd |
CI/CD pipelines across GitHub Actions, GitLab, Jenkins |
loom-code-migration |
Safe code migrations, framework upgrades, codemods |
loom-concurrency |
Concurrency/async patterns across Rust, Python, TS, Go |
loom-crossplane |
Crossplane infra-as-code on Kubernetes APIs |
loom-database-design |
Schema/data model design: relational, NoSQL, warehouse |
loom-data-validation |
Schema validation, input sanitization, output encoding |
loom-data-visualization |
Charts, dashboards, and reports across domains |
loom-dependency-scan |
Scans dependencies for CVEs, license issues, SBOM |
loom-diagramming |
Mermaid diagrams: architecture, sequence, ERD, C4 |
loom-docker |
Dockerfiles, compose, multi-stage builds, hardening |
loom-documentation |
READMEs, architecture docs, changelogs, ADRs |
loom-e2e-testing |
E2E testing with Playwright, Cypress, Selenium |
loom-error-handling |
Error handling: Result/Option, retries, circuit breakers |
loom-event-driven |
Event-driven: queues, pub/sub, event sourcing, sagas |
loom-feature-flags |
Feature flags: rollouts, A/B tests, kill switches |
loom-fluxcd |
GitOps CD for Kubernetes with Flux CD |
loom-git-workflow |
Git branching, commits, merges, conflict resolution |
loom-golang |
Idiomatic Go: goroutines, channels, testing, modules |
loom-grafana |
Grafana dashboards, panels, LogQL/TraceQL queries |
loom-i18n |
i18n/l10n: translations, locale formatting, RTL |
loom-istio |
Istio service mesh: mTLS, routing, canary, Envoy |
loom-karpenter |
Karpenter node autoscaling and cost optimization |
loom-kubernetes |
K8s manifests, Helm, RBAC, operators, troubleshooting |
loom-kustomize |
Kustomize overlays, patches, ConfigMap/Secret gen |
loom-logging-observability |
Structured logging, tracing, metrics, alerting |
loom-model-evaluation |
ML model evaluation: metrics, CV, drift monitoring |
loom-performance-testing |
Load/perf testing with k6, locust, JMeter, Gatling |
loom-prometheus |
PromQL, scrape configs, alerting/recording rules |
loom-prompt-engineering |
LLM prompt design: system prompts, few-shot, security |
loom-python |
Idiomatic Python: FastAPI, Django, pandas, pytest |
loom-rate-limiting |
Rate limiting: token/leaky bucket, quotas, Redis |
loom-react |
React 19+ SPAs: hooks, Jotai, routing, Bun/Vite/Oxc |
loom-refactoring |
Restructures code without changing behavior |
loom-rust |
Idiomatic Rust: ownership, async/tokio, cargo, serde |
loom-search |
Full-text search: Elasticsearch, OpenSearch, Meilisearch |
loom-security-audit |
Deep security audits: OWASP, compliance, hardening |
loom-security-scan |
Quick security checks: secrets, deps, containers |
loom-serialization |
Serialization: JSON, YAML, protobuf, schema evolution |
loom-sql-optimization |
SQL query optimization, indexing, EXPLAIN analysis |
loom-technical-writing |
Technical writing: READMEs, guides, changelogs |
loom-terraform |
Terraform/OpenTofu IaC, modules, state, workspaces |
loom-testing |
Test implementation: unit, integration, e2e, TDD/BDD |
loom-test-strategy |
Test strategy: pyramid, coverage, flaky diagnosis |
loom-threat-model |
Threat modeling: STRIDE, DREAD, PASTA, attack trees |
loom-typescript |
Type-safe TypeScript: generics, strict mode, zod/trpc |
loom-webhooks |
Webhooks: HMAC verification, retries, idempotency |
Note
The catalogued file is the complete skill, not a summary of it — reading it is equivalent to having loaded the skill directly.