A structured workflow system for Claude Code — from idea capture through build, review, and wrap-up, tracked as a unified work record on GitHub Issues. Includes an LLM-as-judge recurring code-health sweep, browser automation, and QA pipeline.
Use when triage or dispatch need a content-aware trust verdict instead of a mechanical label lookup — grant-check informs triage's recommendation, merge-check replaces dispatch's blast-radius gate, failure-check replaces dispatch's blanket failure-revocation rule. Inline helper, never invoked directly by a human. Keywords - autonomy, trust, judgment, grant recommendation, auto-merge, blast radius, failure classification.
Use for browser automation via agent-browser — defines session naming, screenshot/trace paths, and operation vocabulary used by /stories, /visual-review, and /review. Keywords - browse, browser, agent-browser, screenshot, scrape, automation.
Use when implementing a work record, spec, or design doc end-to-end. Accepts a record reference (#N) or spec number (legacy alias) for full lifecycle tracking, or a design doc path to skip /claude-tweaks:specify and build directly from brainstorming output.
Use when capturing ideas that need specification later — brain dumps, half-formed features, things to not forget
Use when you need to challenge assumptions and remove bias from a problem statement before brainstorming. Takes a backlog record or topic and produces a debiased problem framing.
Executes bash commands
Hook triggers when Bash tool is used
Modifies files
Hook triggers on file write and edit operations
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Executables (bin/) — files in this plugin's bin/directory are added to the Bash tool's PATH while the plugin is enabled.
Uses power tools
Uses Bash, Write, or Edit tools
Uses power tools
Uses Bash, Write, or Edit tools
A structured workflow system for Claude Code — from idea capture through build, review, and wrap-up.
Claude Code is powerful but unstructured. claude-tweaks adds a complete development lifecycle: capture ideas, challenge assumptions, decompose into specs, build with quality gates, and learn from what was built. Every finding is explicitly resolved — nothing silently drops.
/claude-tweaks:demo)A new seventh work-record axis (demo:pending / demo:approved / demo:changes-requested)
closes the gap between tests passing, spec completion, and an actual human verifying a built
feature does what was asked. /claude-tweaks:wrap-up applies demo:pending and writes a
Verification Brief (what changed, why, how to verify) while it still has full build context;
the new /claude-tweaks:demo skill aggregates every pending record — across parallel threads,
regardless of merge timing — and captures your verdict.
Every captured idea, health-skill finding, and human-filed issue is now the same thing: one durable work record (a GitHub issue, or its local-files twin — a plain markdown file), tracked through a single spine instead of the old two-file backlog design and per-artifact frontmatter:
BACKLOG ──/specify shapes──► READY ──human grants──► AUTHORIZED ──/dispatch claims──► BUILDING ──user merges──► CLOSED
with parked (on hold, wakes on a trigger) and not-planned (wontfix/duplicate/absorbed) exits at any stage. Two storage drivers back the same taxonomy — work-backend: github-issues (labels + native Issue Types) or work-backend: local-files (frontmatter on a tracked file) — set once by /init, read identically by every consumer skill. See "Work Records" below and skills/_shared/work-record.md for the full contract.
Human-granted auto:build/auto:merge labels replace the retired tier:approved/tier:fast-track/tier:needs-review three-way split — /claude-tweaks:triage is now the interactive grant gate only. A new skill, /claude-tweaks:dispatch, is the queue consumer: it claims an authorized record's whole file-overlap group and hands it to /flow — the triage dispatch headless subcommand no longer exists.
/claude-tweaks:specify and /claude-tweaks:build//flow now materialize a record reference (#N) into a build-time header + spec-shaped body file rather than requiring a pre-existing numbered spec file — the legacy specs/{n}-*.md path still works as an alias for projects that haven't migrated. /claude-tweaks:tidy and /claude-tweaks:help scan the live record queue directly; the former INBOX scan, Deferred-Work scan, and the separate spec index they used to read are retired.
See "Migrating from 5.x" below if this project still carries pre-6.0 state (live tier:*/status:* labels, specs/backlog/ files, or the old backlog-backend flag name).
/claude-tweaks:visualize replaces the external diagram-design companion-plugin integration introduced in v4.7 with a fully native skill — no separate plugin install. It generates self-contained HTML+SVG diagrams (architecture, flowchart, sequence, state, ER, timeline, swimlane, quadrant, nested, tree, org chart, layers, venn, pyramid), themed from the project's own DESIGN.md tokens (or a neutral default skin when Impeccable isn't set up). An optional D2-backed enhanced rendering path handles diagrams-as-code source generation for types with a native D2 construct. The same three soft-hook call sites — /journeys Step 3.6, /specify Step 2.5d, /review Lens 3i-diagram — now suggest invoking it directly, gated by diagram-suggestions: enabled in CLAUDE.md (renamed from diagram-integration:), written by /init Step 11. Diagrams co-locate with what they illustrate (docs/journeys/, docs/plans/) rather than a single central folder; docs/diagrams/ is the fallback for context-free, direct invocations.
/init gains a new Optional Enhancement step: on a detected frontend project without components.json, it offers to bootstrap shadcn/ui — CLI init, plus wiring shadcn's own first-party MCP server into .mcp.json and installing shadcn's official Skill (skills add shadcn/ui), both of which give Claude Code live project context so it stops guessing at component APIs. Writes a shadcn-integration: enabled | cli-only | disabled flag to CLAUDE.md's ## Design integration section (currently write-only — no other skill reads it yet). See /init Step 12.
npx claudepluginhub thomasholknielsen/claude-tweaks-marketplace --plugin claude-tweaksHooks-based workflow orchestration for solo product development with Claude Code
Requirements-driven development workflow with quality gates for practical feature implementation
Unified spec-to-execution workflow system. Combines spec rigor, autonomous execution, and native task orchestration with parallel agent dispatch. Swarms-ready architecture.
Structured project planning and execution through brainstorm, spec, and build phases across three execution tiers: sequential, delegated sub-agents, and full agent teams
Development workflow plugin suite: exploring (thinking partner), requirement-analysis (design workflow), writing-plans / executing-plans (planning and execution), visual-preview, using-git-worktrees / test-driven-development (engineering discipline), acceptance-qa (all-round acceptance), quick-fix (lightweight bug-fix workflow), anysearch (bundled real-time search, vendored from upstream). / 开发工作流插件集:exploring 探索模式、requirement-analysis 设计工作流、writing-plans/executing-plans 计划与执行、visual-preview 可视化预览、using-git-worktrees/test-driven-development 工程纪律、acceptance-qa 全能验收、quick-fix 轻量修复、anysearch 内置实时搜索(上游 skill 内嵌)。
Spec-driven development workflow system with structured phases: Requirements → Design → Tasks → Implementation