Skip to content

Catalog

github-actions[bot] edited this page Jun 16, 2026 · 13 revisions

Catalog

Every official pack — rules, agents, skills, and guardrails — and the items it contains. Item names match filenames in each pack's llm/<kind>/ directory and are used for local overrides.

Rules

Always-on instructions loaded into every AI session automatically.

bluetemberg-rules-accessibility

Accessibility rules for Bluetemberg — WCAG 2.2 A/AA guidance for semantic HTML, ARIA, keyboard navigation, and more.

Rule Enforces
accessible-forms Every form control must have a programmatic label, clear error identification, and grouped fields where appropriate
aria-only-when-needed ARIA is a fallback for when native HTML falls short — never add redundant or conflicting roles
color-contrast Meet WCAG 2.2 AA contrast ratios; never use color as the only visual signal
focus-management Provide visible focus indicators; move focus intentionally on route change and dialog open/close
keyboard-navigation All interactive elements must be reachable and operable by keyboard alone; no keyboard traps
reduced-motion Honor prefers-reduced-motion; never use motion as the sole affordance for a state change
semantic-html Prefer native HTML elements over div+role soup; semantic elements are accessible by default
text-alternatives Provide descriptive alt text for meaningful images; empty alt for decorative; alternatives for icons and media

bluetemberg-rules-agent-memory

Agent memory rules for Bluetemberg — statelessness as a hard constraint, memory provenance, promotion through consolidation, and authority-ranked recall.

Rule Enforces
memory-architecture-checklist Treat LLM statelessness as a hard constraint — decide explicitly what state survives a turn, which storage tier holds it, and how it is evicted.
memory-promotion Session observations never write directly to durable memory — promotion runs through a reviewable consolidation step that dedupes, supersedes, and ratifies.
memory-provenance Every durable memory carries provenance — source, time, confidence, and whether it was user-stated or agent-inferred. No silent writes.
memory-recall-authority Rank recalled memory by authority, not only relevance — and never let stale, contradicted, or superseded memory present as a current constraint.

bluetemberg-rules-context-engineering

Context engineering rules for Bluetemberg — structuring LLM context, managing token budgets, avoiding context pollution.

Rule Enforces
context-pollution-prevention Prevent context pollution — scope tool calls to what you need, don't inject noise into the conversation.
context-positioning Place load-bearing content at the edges of the context, not the middle — models access the start and end far better than the middle.
context-window-budget Manage context window budget explicitly — prune stale content, compress retrieved data, never let context grow unbounded.
multi-turn-context-hygiene Don't keep talking to a derailed conversation — consolidate the requirements and restart fresh. Multi-turn degradation is large and doesn't self-correct.
multi-turn-state-management Manage state across multi-turn interactions — track what was decided, what changed, and what is still open.
prompt-structure Structure prompts into labeled, delimited sections — separate instructions from data, and keep persistent rules in the system message.

bluetemberg-rules-design-craft

Anti-stock design-craft principles for building UI with AI — banned moves, real references, tokens-first, every state.

Rule Enforces
anti-stock-defaults Decide the visual direction before prompting; AI averages toward stock UI, so name what you want and refuse the safe default.
banned-moves-first Keep a concrete banned-moves list and apply it to every UI build; closing off easy defaults forces something more intentional out.
design-every-state Design and build every screen state — empty, loading, error, success — with real content, never just the default with lorem ipsum.
references-not-moods Anchor visual direction to specific named references, not adjectives; concrete coordinates beat moods the model will average away.
tokens-before-components Lock the visual system (palette, type, spacing) as tokens before building UI, and re-state them every refinement to prevent token drift.

bluetemberg-rules-devops

Infrastructure rules for Bluetemberg — Docker, Terraform, Ansible, Kubernetes, Helm, CI/CD, shell scripts, and idempotency.

Rule Enforces
ansible-conventions Follow Ansible best practices for idempotent, maintainable infrastructure automation.
ci-workflow-conventions Secure CI/CD workflows with minimal permissions, pinned actions, and OIDC authentication.
container-image-pinning Pin container image versions to prevent unexpected breaking changes.
docker-best-practices Follow container best practices for secure, efficient Docker images.
helm-conventions Write safe, reusable Helm charts with secure defaults and consistent values structure.
idempotency Every operation — task, script, or deployment — must be safe to re-run without side effects.
jinja2-templates Write safe, readable Jinja2 templates for Ansible and configuration generation.
kubernetes-manifests Write safe, production-grade Kubernetes manifests with resource limits, health checks, and security context.
runbook-discipline Keep operational runbooks and ADRs accurate and commit them alongside the changes they document.
shell-script-standards Write safe, portable shell scripts with error handling and clear control flow.
terraform-conventions Follow Terraform module structure and naming conventions.

bluetemberg-rules-docs

Documentation and diagnostics rules for Bluetemberg — docs parity, post-edit diagnostics, Mermaid diagrams.

Rule Enforces
docs-parity Keep documentation in sync with every user-facing change.
mermaid-diagrams Use Mermaid diagrams instead of ASCII art.
post-edit-diagnostics Run diagnostics and formatter after editing code files.

bluetemberg-rules-git

Git workflow rules for Bluetemberg — branch protection, naming, PR workflow, git move, pre-commit checks.

Rule Enforces
git-move Use git mv for tracked files to preserve history.
git-workflow Branch protection, branch naming, and PR workflow rules.
pre-commit-checks Run formatter, linter, and build checks before every commit.

bluetemberg-rules-llm-api-product

LLM API and product-engineering rules for Bluetemberg — streaming, cost accounting, and cost-aware model selection.

Rule Enforces
cost-accounting Measure LLM cost from the authoritative usage object and the provider's cost API — never from your own token estimates.
model-selection-and-routing Pick models on the accuracy-vs-cost frontier and route easy queries to cheaper models — the most expensive model is rarely the right default.
streaming Stream for perceived latency, but buffer tool-call JSON until the block closes and don't gate safety on partial output.

bluetemberg-rules-nextjs

Next.js rules for Bluetemberg — NEXT_PUBLIC_* env var safety.

Rule Enforces
nextjs-data-fetching Prefer fetch with cache options in Server Components over useEffect + API calls
nextjs-image-optimization Always use next/image; never raw img tags
nextjs-metadata Define metadata/generateMetadata per page; never use in App Router
nextjs-public-env-vars NEXT_PUBLIC_* are inlined into the bundle at build time — never use for secrets or per-environment values.
nextjs-server-components Default to React Server Components; use 'use client' only at leaf interactive nodes

bluetemberg-rules-security

Security guardrail rules for Bluetemberg — secrets management, environment file protection, API error handling.

Rule Enforces
api-error-handling Use structured error responses and never leak internal details.
llm-package-hallucination Verify every LLM-suggested dependency exists in the registry before installing — hallucinated package names are a real supply-chain attack surface (slopsquatting).
never-read-env Never read .env files directly in code.
security-secrets Never hardcode secrets, tokens, or credentials in source code.

bluetemberg-rules-typescript

TypeScript code quality rules for Bluetemberg — type safety, coding standards, early returns, no console.log, design system reuse.

Rule Enforces
coding-standards Keep functions and components small, readable, and easy to reason about.
design-system-reuse Reuse existing shared UI components and design tokens before creating new ones.
early-returns Prefer early returns and guard clauses over nested conditionals.
no-console-log Forbid console.log in production code; use a logger instead.
type-safety Enforce strict type safety — no implicit any, no unguarded assertions.

Agents

Specialist agents invoked for focused, domain-specific tasks.

bluetemberg-agents-a11y-specialist

Accessibility specialist agent for Bluetemberg — WCAG 2.2 A/AA audit and remediation.

Agent Description
a11y-specialist Audits WCAG 2.2 A/AA accessibility — semantic HTML, keyboard, focus, contrast, alt text — and reports prioritized fixes. Use proactively for UI components, forms, dialogs, screen-reader review.

bluetemberg-agents-agentic-specialist

Agentic systems specialist agent for Bluetemberg — agent memory design, state management, orchestration patterns.

Agent Description
agentic-specialist Designs and implements agentic memory, state machines, sub-agent orchestration, and tool-use contracts. Use proactively when building agent loops, persistent memory, handoffs, or tool-call recovery.

bluetemberg-agents-ansible-specialist

Ansible specialist agent for Bluetemberg — roles, playbooks, and Jinja2 templates.

Agent Description
ansible-specialist Writes, reviews, and lints Ansible roles, playbooks, and Jinja2 templates for idempotent infra automation. Use proactively when editing anything under roles/, playbooks, ansible.cfg, or inventory.

bluetemberg-agents-backend-specialist

Backend specialist agent for Bluetemberg — API design, database patterns, error handling, auth.

Agent Description
backend-specialist Builds server-side services — REST/GraphQL APIs, database access, auth flows (JWT, RBAC, OAuth/PKCE). Use proactively for API contracts, DB layers, or server auth. Not CMS schemas or security audits.

bluetemberg-agents-code-reviewer

Code reviewer agent for Bluetemberg — PR review covering patterns, naming, complexity, tests.

Agent Description
code-reviewer Reviews diffs and pull requests for correctness bugs, edge cases, naming, test coverage, and maintainability. Use proactively after code changes. For deep security audits route to security-specialist.

bluetemberg-agents-design-engineer

Design engineer agent for Bluetemberg — builds UI to a reference section-by-section, holds tokens and banned moves, catches stock drift.

Agent Description
design-engineer Builds UI from a Figma comp, screenshot, or visual direction section-by-section, holding design tokens and banned moves. Use when turning a design reference into distinct UI or catching stock drift.

bluetemberg-agents-devops-specialist

DevOps specialist agent for Bluetemberg — CI/CD pipelines, container optimization, IaC review.

Agent Description
devops-specialist Designs and optimizes CI/CD pipelines, Docker images, Terraform IaC, deploy safety, and shell scripts. Use proactively when editing .github/workflows, Dockerfile, terraform/, or Makefile.

bluetemberg-agents-docs-maintainer

Docs maintainer agent for Bluetemberg — documentation synchronization with code changes.

Agent Description
docs-maintainer Updates and audits documentation (README, runbooks, API docs, Mermaid diagrams) to match code, config, and workflow changes. Use proactively after behavior, CLI, API, or schema changes.

bluetemberg-agents-frontend-specialist

Frontend specialist agent for Bluetemberg — UI implementation, design-system, i18n, accessibility.

Agent Description
frontend-specialist Implements and refactors React/Next.js UI — components, state, hooks, SSR/hydration, performance, accessibility. Use proactively when building or changing TSX/JSX UI, not for review or backend work.

bluetemberg-agents-infrastructure-specialist

Infrastructure specialist agent for Bluetemberg — build, CI, container, deployment config.

Agent Description
infrastructure-specialist Maintains build tooling, Dockerfiles, Compose, and dependency manifests/lockfiles for reproducible, secure builds. Use proactively when editing Dockerfile, docker-compose, package.json, or lockfiles.

bluetemberg-agents-kubernetes-specialist

Kubernetes specialist agent for Bluetemberg — manifests, Helm charts, Kustomize overlays.

Agent Description
kubernetes-specialist Writes and reviews Kubernetes manifests, Helm charts, Kustomize overlays — RBAC, probes, resources, security contexts, autoscaling, zero-downtime rollouts. Use proactively when editing k8s YAML.

bluetemberg-agents-security-specialist

Security specialist agent for Bluetemberg — vulnerability audit, dependency scanning, secrets.

Agent Description
security-specialist Audits code for OWASP Top 10 vulns, injection, broken auth/access control, secrets exposure, and dependency/supply-chain CVEs. MUST BE USED for deep security review. Reports findings, does not edit.

bluetemberg-agents-sre-specialist

SRE specialist agent for Bluetemberg — SLOs, alerting, runbooks, and post-mortems.

Agent Description
sre-specialist Reviews production reliability — SLOs, SLIs, error budgets, alerts, runbooks, post-mortems. Use proactively for on-call, observability, or incident-readiness review, not code or security audits.

bluetemberg-agents-test-specialist

Test specialist agent for Bluetemberg — test creation, refactoring, and stabilization.

Agent Description
test-specialist Writes, refactors, and stabilizes unit/integration/e2e tests, fixes flaky/non-deterministic tests, applies fakes and data builders. Use proactively when adding tests or diagnosing flakiness.

Skills

On-demand workflows triggered by slash commands.

bluetemberg-skills-api-design

API design skill for Bluetemberg — RESTful conventions, pagination, versioning.

Skill Description
api-design Designs and reviews RESTful HTTP API endpoints — resource naming, pagination, error contracts, versioning. Use when creating an endpoint, reviewing an API contract, or vetting a breaking change.

bluetemberg-skills-ci-cd-best-practices

CI/CD best practices skill for Bluetemberg — pipeline optimization and caching strategies.

Skill Description
ci-cd-best-practices Audits CI/CD pipeline config (GitHub Actions, GitLab CI) for SHA pinning, dep caching, deploy gates, parallelism, OIDC. Use when adding or editing workflow files or fixing slow/flaky pipelines.

bluetemberg-skills-code-review

Code review skill for Bluetemberg — structured review checklist for PRs.

Skill Description
code-review Reviews a pull request or code diff before merge — establishes intent, reviews changed lines, reports severity-tiered findings with fixes. Use for PR review or pre-PR self-review.

bluetemberg-skills-config-echo

Config echo skill for Bluetemberg — verifies that synced configuration was actually loaded by the LLM.

Skill Description
config-echo Verifies bluetemberg config (rules, agents, skills) is loaded in the session by recalling from context and cross-referencing disk. Use after bluetemberg sync or when a rule seems ignored.

bluetemberg-skills-create-agent

Meta-skill for scaffolding new bluetemberg agents — frontmatter, scoped tools, responsibilities and constraints, and sync.

Skill Description
create-agent Scaffold a new bluetemberg agent in the correct format — frontmatter with scoped tools, responsibilities, constraints, and sync.

bluetemberg-skills-create-pack

Meta-skill for scaffolding a whole new bluetemberg pack — package.json metadata, llm/ layout, catalog registration, and validation.

Skill Description
create-pack Scaffold a whole new bluetemberg pack — package.json metadata, llm/ layout, catalog registration, and validation.

bluetemberg-skills-create-rule

Meta-skill for scaffolding new bluetemberg rules — frontmatter, the Why/Rules/BAD-GOOD/config/gotchas structure, and sync.

Skill Description
create-rule Scaffold a new bluetemberg rule in the correct format — frontmatter, a Why/Rules/BAD-GOOD/config/gotchas body, and sync.

bluetemberg-skills-create-skill

Meta-skill for scaffolding new bluetemberg skills — frontmatter, five required sections, and sync.

Skill Description
create-skill Scaffold a new bluetemberg skill to the deep-skill standard — Protocol, decision trees, BAD/GOOD examples, and a completion checklist.

bluetemberg-skills-design-critique

Design critique skill for Bluetemberg — multi-lens review, hostile QA, and an impact-ranked fix list for built UI.

Skill Description
design-critique Critiques built UI across accessibility, hierarchy, copy, consistency, and device lenses, runs hostile QA, and returns an impact-ranked fix list. Use before shipping a view or when it feels off.

bluetemberg-skills-docs-upkeep

Docs upkeep skill for Bluetemberg — keep docs aligned with implementation changes.

Skill Description
docs-upkeep Updates affected docs in the same task as code or workflow changes — README, API docs, CLI --help, config, wiki. Use when renaming endpoints, adding flags/config, or fixing stale references.

bluetemberg-skills-figma-to-code

Figma-to-code skill for Bluetemberg — the section-by-section design-to-code loop with prompt stacking, drift checks, and surgical refinement.

Skill Description
figma-to-code Translates a Figma comp, screenshot, or mockup into working UI/HTML/CSS code section-by-section. Use when building UI from a design reference or when generated UI drifts from the comp.

bluetemberg-skills-infrastructure-drift-check

Infrastructure drift check skill for Bluetemberg — verify IaC state matches deployed state.

Skill Description
infrastructure-drift-check Detects infrastructure drift before merging IaC changes by diffing declared vs deployed state. Use when reviewing or merging Terraform .tf, Ansible, Kubernetes manifests, Helm, or Compose PRs.

bluetemberg-skills-migration-safety

Migration safety skill for Bluetemberg — database migration review and rollback plans.

Skill Description
migration-safety Reviews or writes database schema migrations against a zero-downtime checklist — destructive ops, lock analysis, batching, rollback path. Use when creating or reviewing a migration pre-deploy.

bluetemberg-skills-patterns

Patterns skill for Bluetemberg — apply reusable architecture and coding patterns.

Skill Description
patterns Finds and reuses an existing codebase pattern instead of inventing new structure. Use when adding files or abstractions, reviewing folder structure, naming, or module boundaries, or refactoring.

bluetemberg-skills-react-patterns

React patterns skill for Bluetemberg — component composition, hook extraction, state co-location.

Skill Description
react-patterns Apply React composition, co-location, hook extraction, and state-placement patterns when writing or reviewing components, hooks, or state. Fixes prop drilling and logic tangled in JSX.

bluetemberg-skills-rollback-plan

Rollback plan skill for Bluetemberg — require tested rollback steps for every production change.

Skill Description
rollback-plan Require and validate a tested rollback plan before merging production deploy, Kubernetes/Terraform, schema-migration, or version-bump PRs. Adds a complexity tier tree, data-loss check, PR template.

bluetemberg-skills-security-audit

Security audit skill for Bluetemberg — dependency audit, secrets scan, OWASP patterns.

Skill Description
security-audit Triages code security findings by severity with detection steps for secrets, injection, auth, input, and dependency risks. Use when reviewing changes pre-deploy or touching auth, input, or uploads.

bluetemberg-skills-stack-change-review

Stack change review skill for Bluetemberg — high-blast-radius infrastructure change review.

Skill Description
stack-change-review Reviews high-blast-radius infra diffs (docker-compose, Ansible group_vars, Terraform, k8s, .env) — port, volume, image, and env risks. Use when reviewing changes to deployed infrastructure.

bluetemberg-skills-sub-agent-design

Sub-agent design skill for Bluetemberg — plan, scope, and implement sub-agent architectures.

Skill Description
sub-agent-design Design a multi-agent system — scope responsibilities, define typed input/output contracts, wire orchestration, add tests. Use when one agent is overloaded or work has parallelizable phases.

bluetemberg-skills-visual-direction

Visual direction skill for Bluetemberg — explore distinct directions, lock one, and produce a banned-moves list and Design System Document.

Skill Description
visual-direction Locks a visual direction before a UI build by exploring 3 options, picking one, writing a banned-moves list + Design System Document. Use when UI keeps coming out generic or before a handoff.

bluetemberg-skills-workspace-hygiene

Workspace hygiene skill for Bluetemberg — clean workspace state during edits.

Skill Description
workspace-hygiene Audits the workspace before a commit or PR — confirms diff scope, flags stray console.log/debugger/scratch files and secrets, checks atomic commits. Use after a refactor or before opening a PR.

Guardrails

Hook-based constraints that fire automatically during AI operations.

bluetemberg-guardrails-git

Git guardrails for Bluetemberg — enforce conventional branch names on AI-created worktrees.

Guardrail Description
conventional-branch-names Enforce conventional branch names before creating a worktree

Overriding a rule

Every rule is a self-contained Markdown file. To change one for your project without forking the pack, create a file with the same name in your local llm/rules/ — it takes priority over the pack version during sync. See Usage for the full priority order.

Clone this wiki locally