ワンクリックで
claude_marketplace
claude_marketplace には amdmax から収集した 78 個の skills があり、リポジトリ単位の職業カバレッジとサイト内 skill 詳細ページを表示します。
このリポジトリの skills
Injects project context (AGENT_DOCS_DIR, active story) before every skill invocation via PreToolUse hook.
Discover skills in this marketplace that match your workflow. Accepts a workload or task description, scans all released skills, and recommends the best matches with install commands. Use when users ask 'what skills are available', 'find a skill for X', 'which skill helps me with Y', 'what can I install', or 'recommend skills for my workflow'.
Collect non-functional requirements through interactive Q&A. Asks targeted questions about performance, scalability, security, reliability, and cost constraints. Invokable with /gather-nfr.
Validates business requirements compliance - generates OpenAPI specs, creates contract tests, and ensures acceptance criteria alignment
Hypothesis-driven bug investigation with mandatory regression tests
Create git commits with configurable numbering (issue-based or sequential). Analyzes staged changes and generates commit messages following project conventions. Optionally creates GitHub issues if no active story exists.
Spawn the Architect agent to map business requirements to implementation details, produce implementation briefs, assess risks, and create ADRs. Use for standalone architecture tasks without the full agile team.
Spawn the Backend Developer agent to implement backend handlers, infrastructure, and configuration to make integration tests pass (TDD green phase). Use for standalone backend implementation without the full agile team.
Spawn the DevOps agent to provision infrastructure (CDK/IaC), schema migrations, and cloud resources. Makes infrastructure tests pass. Use for standalone infra work without the full agile team.
Spawn the Frontend Developer agent to implement frontend code (TypeScript, templates, CSS, build scripts) to make unit and e2e tests pass (TDD green phase). Use for standalone frontend implementation without the full agile team.
Spawn the Test Architect agent to write failing tests (TDD red phase) based on implementation briefs. Tests define contracts for developers to implement against. Use for standalone test authoring without the full agile team.
Fleet orchestrator. Reads implementation-plan.yaml produced by the architect, resolves phase dependencies (topological sort), and spawns specialist agents (devops/backend-dev/frontend-dev) — parallel for independent phases, sequential when dependencies exist. Tracks build state in build-state.yaml.
Issue-tracker-agnostic PR router. Reads $AGENT_DOCS_DIR/project-config.yaml to determine issueTracker (github or gitlab) and delegates to /github:pull-request or /gitlab:pull-request. Defaults to github when config is absent or issueTracker is unset.
PM (Team Lead) — fetch stories, enrich ACs/NFRs, manage story lifecycle, create PRs. Supports story-extract, story-validate, and build subcommands.
Gather technical context from multiple sources including docs, code analysis, architecture files, and user input. Appends comprehensive context to active story for informed decision-making. Invokable with /gather-context.
Pre-implementation scout agent. Analyzes a "Ready for Dev" GitHub story in read-only mode — blast radius, architecture review, ADR check, task list, infra question resolution. Writes YAML report to docs/stories/{id}/. Invokable with /scout:prepare-for-dev [issue_number].
Spin up a 5-agent TDD team (PM, Architect, Test Architect, Backend Dev, Frontend Dev) that fetches the next Ready story and delivers it through a TDD-first workflow. Invokable with /agile-dev-team.
Generate a structured YAML Architecture Decision Record (MADR format) at $ADR_DIR/NNNN-slug.adr.yaml. Validates schema automatically via PostToolUse hook. Use /arch:adr-yaml to produce machine-readable ADR data before rendering with /arch:adr-render.
Generates an AsyncAPI 2.6 spec for new or changed event-driven channels in a story. Reads scout.yaml dependencies for SQS/SNS/EventBridge/Kafka references; writes async-{slug}.asyncapi.yaml to docs/stories/{id}/design/. Skips cleanly if no event channels are detected.
Generate Architecture Decision Records in MADR format with auto-numbering. Analyzes story context, NFRs, and technical context to create comprehensive ADRs. Invokable with /arch:create-adr.
Orchestrates the post-scout design phase. Determines required artefacts, generates ADRs and diagrams, then delegates to arch:openapi-contract, arch:asyncapi-contract, arch:implementation-plan, and arch:design-review. Marks story ready-for-dev once all artefacts are approved.
Posts design artefacts listed in design.yaml to GitHub one at a time, polls the issue for a reviewer reply, handles rejections with clarification comments, and marks the story ready-for-dev when all artefacts are approved. Fully driven through GitHub comments — no re-run required.
Generates enterprise/org-level questions that must be answered before an implementation brief can be safely produced. Covers cloud infrastructure, security & compliance, architecture governance, team ownership, and cost/budget constraints. Skips questions already answered in existing ADRs or the NFR registry. Writes to $STORIES_DIR/{issueNumber}/strategic-questions.yaml (created if absent).
Generates implementation-level questions about conventions and patterns already established in the codebase. Covers integration patterns, data schemas, API conventions, code conventions, testing strategy, and deployment. Every question includes a codebase_hint pointing scout to where the answer likely lives. Writes to $STORIES_DIR/{issueNumber}/tactical-questions.yaml (created if absent).
Produces a phased implementation plan from scout tasks and design artefacts. Mandatory multi-phase when total tasks > 20. Each phase annotated infrastructure|backend|frontend, independently deployable, with a deployability contract. Writes implementation-plan.yaml to docs/stories/{id}/design/.
Generates an OpenAPI 3.1.0 spec for new or changed HTTP endpoints in a story. Reads scout.yaml and relevant handler files; writes api-{slug}.openapi.yaml to docs/stories/{id}/design/. Skips cleanly if no HTTP endpoints are detected.
Comprehensive Claude Code hooks configuration for notifications, validation, auto-formatting, safety checks, and context loading.
Open or file a GitHub issue — bug report, feature request, task, investigation, or test failure. Interactive template-driven workflow with quality checks and validation.
Create or update a PR coupled to the active story in $AGENT_DOCS_DIR/active-story.yaml. Auto-commits uncommitted changes via /git:commit. Checks for an existing open PR on the current branch — creates one if absent, updates it if present. Mirrors labels, assignees, and project from the linked GitHub issue. PR body is minimal: Closes #N plus changed domains only. Auto-resolves merge conflicts after creation.
Quick GitHub issue creation for commit workflow. Creates minimal issues with project prefix and stores in $AGENT_DOCS_DIR/active-story.yaml. Different from /fetch-story (planned work from Projects).
Populate $AGENT_DOCS_DIR/active-story.yaml from a GitHub issue. Accepts an issue number or URL as an optional argument. Falls back to the highest-priority Ready story from GitHub Projects when no argument is provided. Invokable with /fetch-story [number|url].
Close the loop on any completed GitHub issue (story, bug, task, or chore): update the issue body with actual work done, link the PR, apply type and domain labels, assign to self, and add to the GitHub Project board. Use after implementation is merged or verified complete. Accepts a GitHub issue URL or number as argument; falls back to $AGENT_DOCS_DIR/active-story.yaml. Invokable with /gh:story-finalize.
Put a story "in play" by fetching it from GitHub Projects, gathering NFRs, collecting context, and creating ADRs as needed. Invokable with /play-story.
Analyze GitHub Project stories for SMART acceptance criteria and NFR coverage. Applies labels and posts structured feedback comments. Can run as a scheduled background assistant or manually via /check-story-quality.
Implement a scope of work end-to-end — writes code, commits, creates a PR, then loops fixing CI failures until all checks pass. Invoke with a natural language description of the work to do. Example: /until-green add validation to the email field in the signup form.
Convert a $ADR_DIR/NNNN-slug.adr.yaml file into a MADR-format Markdown file. Output path is same as input with .md extension. Use after /arch:adr-yaml.
Push local skills to https://github.com/{{SKILLS_REPO_OWNER}}/{{SKILLS_REPO_NAME}} for reuse across projects. Discovers skills, syncs to remote repository, generates index, and reports results. Use when you want to share skills across projects or when user says "sync skills", "push skills to GitHub", or "share skills".
Validate skill files for proper YAML frontmatter format and valid Claude Code skill options
Install JSON or YAML output styles for Claude Code — changes how Claude formats all responses to structured, machine-readable output. Use when the user wants to respond in JSON, output as YAML, build automation pipelines, pipe Claude output to scripts or tools, or switch to structured response mode. Trigger on phrases like "respond in JSON", "output as YAML", "structured output", "machine readable responses", "install output style", "json mode", "yaml mode".
Generate GitHub-compatible Mermaid diagrams. Enforces GitHub rendering constraints (supported types, <br/> for line breaks, label quoting). Invokable with /github:mermaid-diagram or /mermaid.