By shiggydoodah
A small collection of practical coding-agent skills for Claude Code and Codex.
Read-only discussion and consultation mode — answer questions, explain code, weigh trade-offs, and recommend improvements without implementing anything. Use when the user invokes /ask, asks "what does this do?", presents an idea for discussion, requests a recommendation or suggestion, critiques a feature ("I don't like X, can we improve it?"), or wants to think something through without entering plan mode or having the agent change code.
Reviews a specific area of the codebase for security, performance & scalability, code quality, or test coverage. Invoke with a path, e.g. /audit apps/backend/auth.
Shapes a raw, in-your-head idea into a scoped, non-technical concept brief (IDEA.md) through a collaborative one-question-at-a-time interview — the pre-planning step that comes before there is any plan to refine. Use when the user has an unformed idea to flesh out or scope, says "brainstorm", "help me scope/shape this idea", or "I have an idea", or wants a feature/product concept captured before planning. For stress-testing a plan or document that already exists, use grill-me instead.
Breaks an existing plan, PRD, or tech spec into small, paste-ready task tickets — one self-contained prompt per task that drops cleanly into the current context window or a fresh one. Use when the user says "breakdown", "break this down", wants a plan/PRD/spec split into actionable tasks, steps, or tickets, or wants implementation work they can copy-paste one task at a time. Does not write the plan/PRD/spec — those must already exist.
Ultra-compressed communication mode. Cuts token usage ~75% by dropping filler, articles, and pleasantries while keeping full technical accuracy. Use when user invokes /caveman, says "caveman mode", "talk like caveman", "use caveman", "less tokens", or "be brief".
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.
Practical skills for coding agents.
This is my working collection of agent skills for Claude Code and Codex-style workflows. Some are small habits I want agents to follow every time. Others are larger workflows for planning, reviewing, debugging, writing specs, and getting changes ready for review.
The goal is not to be exhaustive. The goal is to make the useful paths easy to reach and hard to forget.
Discussion
ask - Read-only Q&A, idea discussion, and recommendations — no plan mode, no code changes.Feature docs
brainstorm - Shape a raw idea into an IDEA.md concept brief before there's any plan.grill-me - Stress-test an idea and write CONTEXT.md as decisions settle.create-prd - Turn context or an IDEA.md seed into a structured PRD.md.create-tech-spec - Turn PRD.md into TECH_SPEC.md with adaptive sections and a coverage check.Development workflow
use-tdd - Test-driven development with a red-green-refactor loop.run-diagnose - Reproduce, isolate, instrument, and fix bugs without guessing.audit - Read-only review of a specific area for security, performance, code quality, or test coverage, with findings triaged by severity.Git and GitHub
commit-this - Stage and create clear short or detailed Git commits, with an optional print-only preview.create-pr - Create a high-signal draft GitHub PR from the current branch.review-pr-feedback - Triage GitHub PR review feedback before changing code.code-review - Review the current feature branch diff by default, or review a GitHub PR with optional inline comments.Planning
plan-with-docs - Plan mode that ends in a reviewable plan doc or a refine loop instead of immediate implementation.breakdown - Break an existing plan, PRD, or tech spec into small, paste-ready task tickets (one task per file).Codebase context
create-context - Create one compressed CONTEXT.md for a target app, package, or directory.map-context - Generate a compressed root CONTEXT-MAP.md birds-eye index for faster codebase navigation.Utilities
caveman - Ultra-compressed communication mode.create-a-skill - Create new agent skills with the right structure.Install the plugin from a terminal:
claude plugin marketplace add shiggydoodah/Skills
claude plugin install skills@ll-labs
Or run the same flow inside Claude Code:
/plugin marketplace add shiggydoodah/Skills
/plugin install skills@ll-labs
/reload-plugins
/ask
/ask what does the auth middleware do?
/plan-with-docs
/plan-with-docs @docs/features/auth/TECH_SPEC.md
/breakdown
/breakdown @docs/features/billing/tech_spec.md
/use-tdd
/run-diagnose
/audit apps/backend/auth
/caveman
/create-a-skill
/commit-this
/commit-this --short
/commit-this --print
/commit-this --short --print
/create-pr
/create-pr --open
/create-pr .github/templates/documentation-template.md
/review-pr-feedback
/code-review
/code-review --pr https://github.com/org/repo/pull/42
/code-review --inline --pr https://github.com/org/repo/pull/42
/brainstorm
/grill-me
/create-prd
/create-tech-spec
/create-context apps/web
/map-context
A few notes:
npx claudepluginhub shiggydoodah/skills --plugin skillsPersistent file-based planning for AI coding agents. Crash-proof markdown plans (task_plan.md, findings.md, progress.md) that survive context loss and /clear, with an opt-in completion gate and multi-agent shared state. Manus-style. Works with Claude Code, Codex CLI, Cursor, Kiro, OpenCode and 60+ agents via the SKILL.md standard. Includes Arabic, German, Spanish, and Chinese (Simplified and Traditional).
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Permanent coding companion for Claude Code — survives any update. MCP-based terminal pet with ASCII art, stats, reactions, and personality.
Design fluency for frontend development. 1 skill with 23 commands (/impeccable polish, /impeccable audit, /impeccable critique, etc.) and curated anti-pattern detection.
Behavioral guidelines to reduce common LLM coding mistakes, derived from Andrej Karpathy's observations on LLM coding pitfalls
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.