fntune/skills

Claude Code plugin marketplace — specialized agents, slash commands, and workflow automation for developers.

# Claude Code
claude plugin marketplace add fntune/skills
claude plugin install ts-strict@fntune-skills

# Any agent (Claude Code, Codex CLI, Gemini CLI, ...)
npx skills add fntune/skills

Six focused plugins. Install only what you need.

---

Plugins

ts-strict — TypeScript strict mode migration

Migrates a codebase to strict: true and Biome lint rules using an infrastructure-first strategy: fix the type architecture first, then enable the flags — instead of whack-a-mole on 10,000 individual errors.

Phases: Baseline → Audit (11 error categories) → Fix type infrastructure → Enable strict flags + Biome → Clean

What it handles:

• Categorizes type errors by root cause (missing generics, implicit any, unsafe assertions, missing return types, etc.)
• Full-stack type propagation with Drizzle ORM + Zod v4 patterns
• Monorepo-aware (Turborepo, pnpm workspaces)

Triggers: "enable strict", "strict mode", "fix type safety", "tighten types", "noExplicitAny"

---

swarm-audit — Multi-agent codebase review and fix

Discovers issues from the codebase itself, triages them, then fixes — all with parallel agents. You don't define the rules; the agents find the problems.

Pipeline:

1. Review — partitions codebase into domains, spawns parallel reviewer agents per domain
2. Triage — team lead deduplicates and prioritizes findings
3. Test — optionally checks UI pages after changes (browser-tester agent)
4. Fix — parallel task-executor agents fix issues, domain-assigned
5. Verify — type-check, lint, tests

Use for: security sweeps, type safety audits, code quality passes, pre-release reviews

Triggers: "audit the codebase", "swarm audit", "security sweep", "codebase-wide review"

---

orchestrate — Multi-stage refactoring executor

Takes an RCA or plan document as input, resolves stage dependencies, and executes the plan with parallel agents and task tracking.

What it does:

• Reads stage definitions and depends_on edges from a plan doc
• Spawns parallel Explore agents to assess current state per stage
• Creates a TaskCreate tree with dependency tracking (addBlockedBy)
• Launches task-executor agents for incomplete stages
• Verifies with type-check, lint, tests on completion
• Updates the plan doc with status

Use for: large refactors with an existing plan document, phased migrations, multi-PR work

Triggers: implementing a phased plan with stage dependencies

---

design-exploration — Parallel design variants

Spawns N agents in parallel, each producing one self-contained page variant with a distinct aesthetic direction. All run simultaneously — you get N variants in the time it takes to generate one.

Output: each variant is a standalone page file with its own styles and layout, ready to compare side by side.

Triggers: "design exploration", "design variants", "parallel designs", "generate N designs"

---

trace — Dependency tree mapper

Maps codebase dependencies from entry points to leaves. Outputs a dependency tree and a numbered reading order (leaves first, entry last). Uses only Grep, Glob, and Read — no model calls, instant results.

Output:

• Nested dependency tree from entry point to all leaves
• Numbered reading order for onboarding or refactoring planning

Use for: onboarding to a new codebase, understanding what to read before moving files

Triggers: "trace dependencies", "map codebase", "reading order", "dependency tree"

---

utils — Slash commands and subagents for everyday workflows

Nine slash commands and three subagents for common developer tasks.

Slash commands:

Command	What it does
/ask	Structured decision gathering with tradeoff analysis
/commit-push	Commit and push the current branch
/debloat	Remove AI slop and unnecessary code
/dry	Find DRY violations — duplicated logic, repeated patterns
/lint	Run linter and report results
/triage	Review issues against code, identify patterns, plan fixes
/update-claude	Update CLAUDE.md from conversation discoveries
/web-interface-guidelines	Check UI code against Vercel Web Interface Guidelines
/worktree-diff	Compare a file across all worktrees and consolidate changes

Subagents:

Agent	What it does
code-audit-reviewer	Systematic code review against specs; catches AI-introduced issues, missing error handling, and inconsistencies
quick-refactor	Fast mechanical refactoring — renames, moves, reformats, import updates
task-executor	Precise execution of well-defined tasks without deviation or interpretation

---

Installation

Claude Code (native plugin)

# Add the marketplace once
claude plugin marketplace add fntune/skills

# Install individual plugins
claude plugin install ts-strict@fntune-skills
claude plugin install utils@fntune-skills
claude plugin install swarm-audit@fntune-skills
claude plugin install design-exploration@fntune-skills
claude plugin install orchestrate@fntune-skills
claude plugin install trace@fntune-skills

Cross-agent (skills.sh)

Works with Claude Code, Codex CLI, Gemini CLI, and 30+ other agents:

# All skills
npx skills add fntune/skills

# Single skill
npx skills add fntune/skills@ts-strict
npx skills add fntune/skills@swarm-audit

---

Plugin structure

<plugin>/
  .claude-plugin/plugin.json    manifest (name, version, description, triggers)
  skills/<name>/SKILL.md        contextual knowledge — loaded when relevant
  commands/<name>.md            slash commands — user-invokable actions
  agents/<name>.md              subagents — delegated task specialists

---

License

MIT