crafting-system-claude

Crafting Kit architectural decision records covering risk gates, execution as a first class stage, no assumption execution, and tool permission enforcement. Use when you need the rationale behind a contract or boundary.

Audit API, CLI, OpenAPI, GraphQL, and CLI help against documented contracts and example outputs. Use when validating contract drift or releasing public surfaces.

Audit API contract compatibility, deprecation strategy, and version migration. Use when planning a contract change.

Answer specific repository questions with cited paths and excerpts. Use when the user asks where something is, how it works, or what depends on it.

Compare the project to competitors using source evidence and allowed research. Use when preparing positioning, market readiness, or product narrative.

Assess database migration reversibility, idempotency, transactional safety, and destructive-change risk. Use before applying or shipping a migration.

Enforce module and package dependency boundaries against a target architecture document. Use when reviewing imports against architecture rules.

Crafting Kit Mermaid diagrams of the global orchestrator, execution state machine, and tool permission flow. Use when explaining the runtime contract visually.

Crafting Kit example prompt sequences and user messages for codebase analysis, refactor, AI ready repo, scenario decisions, and full prompt usage walkthroughs. Use when teaching how to chain prompts end to end.

Generate an executive summary of repository readiness across security, supply-chain, OSS, observability, performance, docs, compliance, API, data, AI-readiness, MCP, and release. Use when reporting up to leadership.

Crafting Kit failure-mode catalog and implementation-claim audit. Use when classifying a regression or auditing a claim.

Validate generated artifacts against golden output fixtures with exact-match or fuzzy comparison. Use when running prompt regression suites.

Crafting Kit manifests and stage routing including prompt manifest, skill manifest, agent manifest, prompt to skill map, prompt aliases, stage manifest, evidence ledger, source to claim map, and prompt inventory. Use when auditing the prompt skill or agent contract.

Audit logging, metrics, tracing, alerting, and incident playbooks. Use when planning SRE work or production readiness.

Audit performance and capacity readiness including hotpaths, slow tests, unbounded queries or loops, and load-test coverage. Use when reviewing performance posture.

Crafting Kit operational policy YAMLs including tool permission policy, secret handling, hard stop triggers, research policy, risk scoring, execution class matrix, and scheduled scoring. Use when you need the exact policy contract.

Apply a strict pull-request review and merge gate using diffs, tests, and validation truth. Use when deciding whether a branch is ready to merge.

Compress a prompt without weakening contracts, gates, or output schemas. Use when token budget is tight.

Detect contradictions, redundancies, and missing dependencies across a prompt set. Use when auditing a prompt library.

Full Crafting Kit prompt library (132 prompts) covering source analysis, glossary/context, feature inventory, HLD/LLD, ADR/RFC, refactoring families, code review personas, Git/release safety, secret-aware execution, MCP tooling, readiness gates, OSS packaging, and prompt assurance. Use when you need the full text of a specific prompt or stage.

Audit release, launch, and deployment readiness including rollout, rollback, smoke tests, and runbook. Use when preparing a release.

Render a go/no-go release board from readiness gate evidence. Use when calling a release decision.

Build a repository intelligence database (symbols, deps, tests, owners, surfaces) for fast Q&A and audits. Use when standing up a knowledge base for a repo.

Compare runtime surfaces (HTTP, CLI, jobs, events, MCP) against documented or generated descriptors. Use when checking surface drift after refactor.

Crafting Kit YAML schemas for risk score, execution gate, approval request, minimum input request, scheduled run report, and scorecard. Use when generating or validating gate output.

Use to answer high-level and low-level questions about repository structure, runtime surfaces, and source paths.

Use to write README, OSS docs, docs maps, and AI context docs from repository truth.

Use to classify feature status as implemented, partial, docs-only, mock-only, test-only, not found, or unknown.

Use to explain project terms, domain vocabulary, technical jargon, and non-technical meanings from repository evidence.

Use to design safe, behavior-preserving refactoring waves from repository evidence.

Crafting Kit document templates for CLAUDE.md, AGENTS.md, ADR, RFC, project glossary, feature inventory, AI context, question routing map, and reports. Use when generating one of these artifacts.

Scan a Claude or LLM output for unsupported claims, fabrications, or missing evidence. Use when validating a generated report.

Audit operational excellence, cost, sustainability, and well-architected posture. Use when preparing a cost or sustainability review.

Review or design Kubernetes runtime operations including health, readiness, graceful shutdown, resource sizing, security context, and rollout. Use when working with Dockerfile, Kubernetes manifests, Helm, or rollout strategy.

Package, validate, and ship a Claude Code or Cowork plugin. Use when building, packaging, or publishing a plugin manifest, skills, agents, hooks, or marketplace entry.

Generate changelog entries, migration notes, and release notes from real diffs and commits. Use when preparing a release or tagging a version.

Maintain an engineering risk register with severity, likelihood, evidence, and remediation owner. Use when capturing technical risk for a project.

Inspect a repository and produce a source-first artifact pack covering languages, frameworks, conventions, runtime surfaces, build pipeline, tests, and configs. Use when starting work on an unfamiliar repository.

Design a testing strategy and pyramid for a repository or change. Use when planning unit, component, contract, integration, e2e, smoke, fuzz, or benchmark coverage.

Bounded code authoring and editing with source-first analysis, risk and execution gates, and validation truth. Use when the user asks to write, edit, refactor, or fix code in a repository.