Skills from the Artificer
The-Artificer-of-Ciphers-LLC/skills-from-the-artificer
29 skills Claude Code: roteamento de custo, debugging e referência de leis
Ver no GitHubColeção de 29 skills Claude Code prontas para uso: roteamento de modelos por custo, debugging estruturado (rubber-duck, test-first, trust-but-verify) e 24 skills de referência sobre leis de engenharia de software. Instale via CLI ou copie para ~/.claude/skills/.
Instalação
npx convez-ai The-Artificer-of-Ciphers-LLC/skills-from-the-artificer1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Otimizar custo de execução roteando trabalho ao modelo mais barato adequado
- Debugging estruturado com rubber-duck e test-first-bugfix para rastrear bugs
- Validação de claims contra fontes primárias (código, docs, specs)
Veredito

Conjunto robusto e pragmático de skills para equipes que levam debugging e custo a sério. Foco em leis de software e validação torna-o mais acadêmico que a maioria, exigindo adoção cultural para valer.
README
A tradução para o português deste README ainda está na fila.
skills-from-the-artificer
A collection of Claude Code skills, distributed as drop-in directories. Each top-level folder is a self-contained skill — install it with the skills CLI, or copy it into ~/.claude/skills/ and restart Claude Code.
Skills
| Skill | What it does |
|---|---|
skills-from-the-artificer/ | Meta dispatcher. Given a change, fix, diff, design, or decision, classifies it and points you at the relevant law/principle skill(s) from the 24-law collection below. The entry point for "which laws apply here?" and the Bug Remediation Step 2 cross-reference. |
cost-tier-routing/ | Routes work to the cheapest model that can do it correctly — haiku for search/IO, sonnet for coding, opus for orchestration. Ships a skill and three model-pinned subagents. |
cost-routing/ | Top-level dispatcher that classifies requests into scout / coder / architect tiers and ships dispatch templates + a PreToolUse warning hook. Companion form to cost-tier-routing. |
rubber-duck/ | Interactive rubber-duck debugging session. Forces you to reconstruct your mental model of the bug from scratch, which is where the bug usually surfaces. |
test-first-bugfix/ | Test-driven bug fixing — reproduce the bug as a failing test before you touch the fix. Catches "fixes" that don't actually fix anything and prevents regressions. |
trust-but-verify/ | Re-validates every claim a subagent hands back against a primary source — the code, docs/ADRs, the memory dir, context7, or a language spec — before you act on it. Treats a report as a lead, not a fact: nothing is verified until a source was opened and quoted. |
artifact-gates/ | Makes workflow steps enforceable. A step that produces no artifact cannot be enforced, and will be skipped — so each blocking step writes a file, and the next action is denied until that file exists. Ships a config-driven PreToolUse hook, two example contracts, and a 69-case behavioral suite. Turns self-attestation ("I ran the review") and prose policy ("admin-merge only for a missing reviewer") into machine-checked preconditions. |
ci-preflight/ | Pre-push checklist that prevents CI whiplash — multiple red pushes that could have been caught locally. Guards against three anti-patterns: missing registration surfaces when adding new shipped files, skipping npm test before pushing, and guessing at cross-platform fixes instead of diagnosing the root cause. |
gsd-core-hooks/ | A full hook suite, lifted from a real project. Fourteen PreToolUse/PostToolUse scripts, each closing one named, dated incident: a passing test-runner verdict bound to the exact commit sha before push, no local node --test, Memtrace evidence required on a PR approval, and guards against harness poll/timeout false-negatives (backgrounded dispatch, foreground poll-loop kills, self-matching pgrep waits). Backs the rules commands/ states in prose. Not generic — see its README.md for what to adapt. |
writing-docs-with-diataxis/ | Write, restructure, or audit docs with the Diátaxis framework (tutorial / how-to / reference / explanation). Splits mixed READMEs by reader need and includes a mapping for documenting a skills repo. |
Migrated from the Mac mini (Swift/iOS, App Store Connect, ADR coverage)
Some of these originate upstream (e.g. the swiftui-* and asc-* sets); they're collected here so one install covers every machine. The Dimillian/Skills set (swiftui-ui-patterns, swiftui-liquid-glass, swiftui-performance-audit, ios-debugger-agent, …) is deliberately not vendored — install it from upstream.
| Skill | What it does |
|---|---|
adr-phase-coverage/ | Use when a user needs to turn a design doc, ADR, RFC, or spec into a phased implementation plan — or verify that an existing phase… |
apple-on-device-ai/ | Integrate on-device AI using Foundation Models framework, Core ML, and open-source LLM runtimes on Apple Silicon. Covers Foundation Models… |
asc-app-create-ui/ | Create a new App Store Connect app record via browser automation. Use when there is no public API for app creation and you need an agent to… |
asc-build-lifecycle/ | Track build processing, find latest builds, and clean up old builds with asc. Use when managing build retention or waiting on processing. |
asc-cli-usage/ | Guidance for using asc cli in this repo (flags, output formats, pagination, auth, and discovery). Use when asked to run or design asc… |
asc-crash-triage/ | Triage TestFlight crashes, beta feedback, and performance diagnostics using asc. Use when the user asks about TF crashes, TestFlight crash… |
asc-id-resolver/ | Resolve App Store Connect IDs (apps, builds, versions, groups, testers) from human-friendly names using asc. Use when commands require IDs. |
asc-localize-metadata/ | Automatically translate and sync App Store metadata (description, keywords, what's new, subtitle) to multiple languages using LLM… |
asc-metadata-sync/ | Sync and validate App Store metadata and localizations with asc, including legacy metadata format migration. Use when updating metadata or… |
asc-notarization/ | Archive, export, and notarize macOS apps using xcodebuild and asc. Use when you need to prepare a macOS app for distribution outside the… |
asc-ppp-pricing/ | Set territory-specific pricing for subscriptions and in-app purchases using current asc setup, pricing summary, price import, and price… |
asc-release-flow/ | End-to-end release workflows for TestFlight and App Store using asc publish, builds, versions, and submit commands. Use when asked to… |
asc-revenuecat-catalog-sync/ | Reconcile App Store Connect subscriptions and in-app purchases with RevenueCat products, entitlements, offerings, and packages using asc… |
asc-shots-pipeline/ | Orchestrate iOS screenshot automation with xcodebuild/simctl for build-run, AXe for UI actions, JSON settings and plan files, Go-based… |
asc-signing-setup/ | Set up bundle IDs, capabilities, signing certificates, and provisioning profiles with the asc cli. Use when onboarding a new app or… |
asc-submission-health/ | Preflight App Store submissions, submit builds, and monitor review status with asc. Use when shipping or troubleshooting review submissions. |
asc-subscription-localization/ | Bulk-localize subscription and in-app purchase display names across all App Store locales using asc. Use when you want to fill in… |
asc-testflight-orchestration/ | Orchestrate TestFlight distribution, groups, testers, and What to Test notes using asc. Use when rolling out betas. |
asc-wall-submit/ | Submit or update a Wall of Apps entry in the App-Store-Connect-CLI repository using the existing generate-and-PR flow. Use when the user… |
asc-workflow/ | Define, validate, and run repo-local multi-step automations with asc workflow and .asc/workflow.json. Use when migrating from lane… |
asc-xcode-build/ | Build, archive, and export iOS/macOS apps with xcodebuild before uploading to App Store Connect. Use when you need to create an IPA or PKG… |
developing-ios-apps/ | Develops iOS/macOS applications with XcodeGen, SwiftUI, and SPM. Handles Apple Developer signing, notarization, and CI/CD pipelines.… |
ios-accessibility/ | Implements, reviews, or improves accessibility in iOS/macOS apps with SwiftUI, UIKit, and AppKit. Use when adding VoiceOver, Voice Control,… |
mobile-ios-design/ | Master iOS Human Interface Guidelines and SwiftUI patterns for building native iOS apps. Use when designing iOS interfaces, implementing… |
swift-concurrency/ | Diagnose Swift Concurrency issues, refactor callback-based code to async/await, and guide Swift 6 migration when working with tasks,… |
swift-mcp-server-generator/ | Generate a complete Model Context Protocol server project in Swift using the official MCP Swift SDK package. |
swift-testing-expert/ | Expert guidance for Swift Testing: test structure, #expect/#require macros, traits and tags, parameterized tests, test plans, parallel… |
swiftdata/ | Implement, review, or improve data persistence using SwiftData. Use when defining @Model classes with @Attribute, @Relationship,… |
swiftui-expert-skill/ | Use when writing, reviewing, or refactoring SwiftUI code for iOS or macOS, including state management, view composition, performance,… |
swiftui-pro/ | Comprehensively reviews SwiftUI code for best practices on modern APIs, maintainability, and performance. Use when reading, writing, or… |
xcodebuildmcp-cli/ | Official skill for the XcodeBuildMCP CLI. Use when doing iOS/macOS/watchOS/tvOS/visionOS work (build, test, run, debug, log, UI automation). |
test-orphan-guard/ | Reaps orphaned test-runner processes left behind by agent test runs; ships a guarded test wrapper, a Stop-hook entry, and shell integration. |
no-defer/ | Prevents an agent from deferring, postponing, or skipping parts of a task — every item must be completed before reporting done. |
qa-test-architect/ | Risk-first test planning and authoring for any codebase: risk assessment, test-case design (happy / boundary / negative / independence), then implementation in the project's own stack with a verified run. |
swift-agents/ | Swift agent team (fork of Techopolis/swift-agents, MIT). swift-lead orchestrator plus 15 specialist subagents (concurrency, SwiftUI, SwiftData, StoreKit, accessibility, on-device AI, …) and a hook that delegates Swift tasks to the lead. Installs to ~/.claude/agents/, not skills/; see its README. Upstream last updated March 2026; maintained here. Fork: The-Artificer-of-Ciphers-LLC/swift-agents. |
Debugging, release and money-handling workflow (generalized from a production iOS project)
| Skill | What it does |
|---|---|
fix-regression-guard/ | Mandatory regression analysis before committing a bug fix. Use before any fix commit, after any change to a guard, condition, ownership check, or… |
post-fix-trace/ | Second-pass regression trace after a bug fix, to catch what the fix itself broke. Use after completing a fix from your debug workflow (trace-based… |
debug-test-capture/ | Turns a finished debug session into permanent regression tests. Use right after a bug fix is verified or committed: after a post-fix regression trace… |
trace-fix/ | Evidence-gated debugging protocol for iOS/macOS and Swift apps: trace the execution path, locate the exact divergence, prove the root cause, and only… |
work-journal/ | Record and recall per-branch session decisions across sessions and agent worktrees using an append-only local markdown journal. Use at the end of… |
testflight-publish/ | Orchestrates an end-to-end TestFlight release: pre-flight checks, archive, upload to App Store Connect, waiting for build processing, distributing to… |
testflight-release-notes/ | Writes tester-facing TestFlight "What to Test" notes and App Store "What's New" release notes from raw developer input such as a git log since the… |
money-boundary-judgment/ | How to cross an API that forces Double (framework parameters, chart plot values, any Double-only library) without breaking a "Decimal for money"… |
ios-simulator-boot-health/ | Diagnose and handle iOS simulator boot failures ("PosterBoard status -5", no healthy simulator, a shard that ran no tests) on a shared Mac: what the failure is, safe vs unsafe simctl commands, and judging a test run by its log and xcresult instead of an exit status. Written for Coincert's scripts/test-ios.sh; adapt the UDIDs and env vars. |
Not a skill
| Directory | What it is |
|---|---|
claude-guards/ | Personal guard hooks, version-controlled. Thirteen PreToolUse/SessionStart/Stop/SubagentStop scripts kept here and symlinked into ~/.claude/hooks/, so the canonical copy is in git rather than only on one machine. Registered at user scope, so they apply across every project: Memtrace-first code discovery, model-tier routing, no-defer enforcement, worktree discipline, CI-rerun blocking. Three ship regression suites (199 / 49 / 66 cases). Not packaged, no SKILL.md, not installable with the skills CLI — read it as a worked example of hook scoping, not as a drop-in. |
Distinct from gsd-core-hooks/, which is a packaged skill documenting one
project's suite. claude-guards/ is cross-project policy.
Commands
commands/ holds workflow directives as Claude Code slash commands —
/feature-builder, /bug-fixer, /triage-review, /review-open-prs. These replace the former
directives/ folder: same procedures, now invocable rather than pasted, with frontmatter declaring
the tools each may touch.
cp commands/*.md ~/.claude/commands/
They are written against one project's toolchain (its test runner, label vocabulary, and controlling
files) and are a starting shape rather than a drop-in — see commands/README.md
for what to adapt. Three of them depend on artifact-gates/ for enforcement;
without that hook their artifact contracts are just prose. gsd-core-hooks/ ships
the fuller suite these commands were originally run against — test-verdict gating, no-local-test,
Memtrace-first — as a reference for what "machine-enforced" looked like in the source project.
Laws of software (24 reference skills)
A pack of software-engineering laws, each loaded as an on-demand reference skill. Every law is its own top-level directory:
choose-boring-technology/ ·
conways-law/ ·
cunninghams-law/ ·
doerrs-law/ ·
fitts-law/ ·
galls-law/ ·
goodharts-law/ ·
greenspuns-tenth-rule/ ·
hofstadters-law/ ·
hyrums-law/ ·
kerckhoffs-principle/ ·
kernighans-law/ ·
knuths-optimization-principle/ ·
lady-lovelaces-objection/ ·
leaky-abstractions/ ·
linuss-law/ ·
moores-law/ ·
norvigs-law/ ·
parkinsons-law/ ·
peter-principle/ ·
postels-law/ ·
shirky-principle/ ·
wirths-law/ ·
zawinskis-law/
Book-derived design skills
Review and audit skills distilled from A Philosophy of Software Design and Domain-Driven Design. Each reports findings and hands off to a neighboring skill for redesign or documentation:
aposd-red-flag-review/ ·
ddd-strategic-audit/ ·
ddd-tactical-review/
Install (recommended: the skills CLI)
The skills CLI discovers each top-level skill and installs it for you. Add -g to install globally (user-level, into ~/.claude/skills/) instead of into the current project.
# install everything
npx skills add The-Artificer-of-Ciphers-LLC/skills-from-the-artificer --all -g
# or choose interactively
npx skills add The-Artificer-of-Ciphers-LLC/skills-from-the-artificer -g
# or install specific skills
npx skills add The-Artificer-of-Ciphers-LLC/skills-from-the-artificer -g --skill conways-law,rubber-duck
Install (manual)
Each skill is also a cp -r away:
# pick the ones you want
cp -r rubber-duck ~/.claude/skills/
cp -r test-first-bugfix ~/.claude/skills/
cp -r conways-law ~/.claude/skills/
# or grab every law at once
cp -r *-law *-principle *-rule *-objection *-technology *-abstractions ~/.claude/skills/
Restart Claude Code. The skill discovery scan runs at session start.
Install (skills that ship agents too)
cost-tier-routing/ is unusual — it includes companion subagents that have to land in ~/.claude/agents/, not ~/.claude/skills/. Its README.md documents the two-step install. The skill alone is inert without the agents because the model: pin lives in the agent files.
cd cost-tier-routing
mkdir -p ~/.claude/skills/cost-tier-routing ~/.claude/agents
cp SKILL.md ~/.claude/skills/cost-tier-routing/
cp agents/*.md ~/.claude/agents/
Install (skills that ship hooks too)
cost-routing/, artifact-gates/, and gsd-core-hooks/ include hooks that must land in
~/.claude/hooks/ and be registered under PreToolUse/PostToolUse in settings.json. Each
has a README.md with the exact snippet.
An unregistered hook — or one registered with a path your Claude Code version does not expand —
fails open silently, and that looks exactly like a hook that is working and has nothing to
complain about. artifact-gates/ ships test/gate-test.sh so you can prove the engine works before
trusting it; confirm the wiring separately by triggering a gate on purpose.
How skills work
A skill is a directory containing a SKILL.md with YAML frontmatter — name and description — followed by the prose Claude Code reads when the skill triggers. The description is the trigger surface: Claude scans loaded skill descriptions against the current task and invokes a skill via the Skill tool when the description matches. Skills can ship supporting files (eval suites, reference scripts, sub-skills) alongside SKILL.md.
See the Claude Code skills docs for the canonical reference.
Evals
Some skills include an evals/evals.json file with prompts and expected behaviors. These are consumed by the skill-creator workflow to benchmark whether a skill triggers correctly and produces the expected routing decisions. The format is intentionally minimal:
{
"skill_name": "<name>",
"evals": [
{
"id": 0,
"prompt": "<user prompt that should trigger the skill>",
"expected_output": "<what Claude should do, prose>",
"files": []
}
]
}
Skill authoring conventions
Every SKILL.md here follows three rules (re-audited November 2026 against Anthropic's updated guidance for the Claude 5.5 models):
- The first 100 lines carry the skill. Trigger context, safety rules, "agent behavior" sections and the workflow outline come first, and a
SKILL.mdstays under 500 lines. Bulky detail (templates, long scripts, project-specific pitfalls, API catalogs) lives inreferences/*.md, linked fromSKILL.mdwith a sentence saying when to read it, and every reference file longer than about 100 lines opens with a## Contentslist so the model can jump to a section instead of reading the file whole. - Match prescriptiveness to the step. Low freedom (exact commands, required output shapes, hard rules) where a mistake is expensive or irreversible: uploads, deletes, merges, money, accessibility requirements, evidence gates. Medium freedom (a goal plus boundaries or a template) for most workflow steps. High freedom for synthesis and judgment. A rule earns a MUST only when it closes a named failure; otherwise state the goal and let the model pick the method.
- Multi-step workflows carry a progress checklist and a feedback loop. A
- [ ]list near the top names each step and what proves it done, and any step with a validator says "fix what it found, then run it again" instead of running once.
The law and principle skills (about 60 lines each) are reference lenses rather than workflows and only need rule 1.
Contributing a skill
- Create a top-level directory named after the skill (kebab-case).
- Drop in a
SKILL.mdwith the frontmattername:anddescription:fields, followed by the body Claude should read, laid out per the authoring conventions above. - Optionally add
evals/evals.jsonwith trigger prompts and expected behaviors. - Optionally add a per-skill
README.mddocumenting anything unusual about the install (e.g. companion agents). - PR.
License
Skills in this repo are MIT-licensed unless a skill's own directory says otherwise.