skill-madness

Plugin skill-madness from marketplace skill-madness in ivy00johns/Skill-Madness. Commands read from the repo's manifest — 2 days ago.

Skills (87)

• git-branch-cleanup

  Clean up stale and merged git branches — both local and remote. Prunes tracking refs, identifies merged branches safe to delete, flags orphaned branches whose remote is gone, and presents a clear summary before acting. Use this skill whenever the user mentions cleaning up branches, pruning stale bra

• git-clean-worktrees

  Clean up stale git worktrees whose work is already on the main branch. Worktrees accumulate after branches are merged — they sit around with "modified" files that look like uncommitted work but are actually already on main. This causes confusion ("did I forget to merge this?") and clutters git statu

• skill-audit

  Scan all skills (or a filtered subset) for consistency, quality issues, gaps, and ownership conflicts in bulk. Use this skill when auditing the full skill ecosystem, running a broad quality scan, checking for ownership overlaps across agents, finding inconsistencies between skills, doing a "health c

• skill-deep-review

  Perform a thorough, single-skill deep dive reviewing structure, description quality, instruction clarity, progressive disclosure, anti-patterns, and frontmatter compliance — then run test prompts via /skill-creator to validate triggering and output quality. Use this skill when deeply reviewing one s

• skill-improvement-plan

  Consume review reports from skill-deep-review or skill-audit and produce a prioritized, actionable improvement plan with specific edits per skill. Use this skill when you have review feedback to act on, need to plan skill improvements, want to turn audit results into a fix plan, or when someone says

• skill-updater

  Execute skill improvement plans by making edits to SKILL.md files and references, validating changes, and optionally syncing to global skill locations. Use this skill when implementing planned skill changes, applying fixes from a review, executing an improvement plan, batch-editing skills, or when s

• contract-auditor

  Audit implementations against integration contracts before runtime integration testing, in native orchestrator dispatch or explicitly requested solo static review. Read code/contracts and report mismatches; never run the app or change implementation.

• contract-author

  Generate machine-readable integration contracts (API, data layer, shared types, events) before implementation starts. Use when writing OpenAPI or AsyncAPI specs, Pydantic models, TypeScript interfaces, JSON Schema or data-layer interfaces, or when the orchestrator needs contracts for a plan. Trigger

• git-commit

  Conventional commit format, allowed types and branch naming for this repo. Use when staging changes, writing a commit message, or creating or naming a branch, and whenever you're about to run `git commit`. Trigger on "commit", "commit this", "write a commit message", "new branch", "branch name". dir

• git-post-merge-cleanup

  Clean up stale git state after merges in one pass: merged local branches, tracking refs whose remote is gone, and worktrees that look modified but are already on main. Scans, shows a plan, acts only on confirmation; supports --dry-run and --yes. Use after a batch of PRs lands. Trigger on "clean up b

• git-pr-feedback

  Fetch, triage and address PR review comments from Copilot and human reviewers. Use when asked to check or respond to PR feedback, or when returning to a PR with accumulated comments. Trigger on "check PR comments", "address PR feedback", "fix review comments", "what did copilot say". directories: []

• git-pr

  PR title format, body structure and the gh CLI workflow for creating or updating pull requests in this repo. Use when opening a PR, writing or updating a PR description, or preparing changes for review, and whenever you're about to run `gh pr create`. Trigger on "create a PR", "open a PR", "PR descr

• babysit

  Keep an open PR healthy on a schedule: poll for review activity, auto-address bot and Copilot nits and routine rebases, re-run the gate, until the PR is mergeable. HITL on anything irreversible; never auto-merges. The scheduled loop around git-pr-feedback, run as "/loop 5m /babysit". Trigger on "bab

• codebase-exploration-loop

  Build a mental model of an unfamiliar codebase by looping: read-only subagents map subsystems, you synthesize and find gaps, and repeat until a written architecture summary answers every seed question with code citations. Use when onboarding to a repo or before touching code you don't understand. Tr

• contract-conformance-loop

  Build until every criterion of an authored contract holds: the builder works one failing criterion at a time and a fresh-context evaluator (contract-auditor, no write tools, default-FAIL) grades all of them. Use when a contract or acceptance-criteria spec exists. Trigger on "build until the contract

• coverage-loop

  Grow a test suite to a coverage target: find the lowest-covered meaningful unit, add real behavior tests, re-measure, until coverage meets the target in .claude/profile.yaml and the suite is green. Assertion-free tests and coverage excludes count as findings. Use when coverage is below target. Trigg

• dependency-health-loop

  Keep dependencies healthy on a schedule: audit for vulnerabilities and stale pins, apply one safe update per pass, prove it with the full gate, and open or update a PR. HITL on every major bump; never auto-merges. Use to have deps watched during a sprint. Trigger on "audit my dependencies", "keep de

• fix-until-green

  Drive a failing build to passing: run the full gate (tests, lint, typecheck), fix one root cause, re-run everything, until all three exit clean. Hard cap, no-progress detection, no cheating the gate. Use when asked to drive red tests or CI green, or as the QE inner loop; not merely because a test fa

• loop-controller

  The harness every loop skill builds on: a verifiable stop condition, the mandatory guardrails (iteration cap, enforced budget, no-progress detection), and the right primitive (/goal, /loop, Stop hook, bash Ralph loop, workflow). Use when work must continue until something is provably true, or to aut

• migration-loop

  Migrate an enumerated set of targets one per iteration: transform, verify the suite stays green, checkpoint, mark done, until every target is migrated and a grep for the old pattern returns zero. Use for codemods, framework or test-runner swaps, API version bumps, renames across a repo. Trigger on "

• nightly-docs-and-changelog

  A nightly cloud routine that keeps an existing repo's docs and CHANGELOG.md in step with the code: diff since the last run, fix docs for what changed, log user-relevant changes, open one PR for review, never merge. Survives laptop-off. Use when docs keep drifting or the changelog is stale. Trigger o

• orchestrator-task-loop

  Outer loop over the Agent Teams shared task list: keep assigning the top unblocked task to idle teammates and gating each completed task until every task passes its TaskCompleted gate. Experimental, behind CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1; falls back to the orchestrator otherwise. Trigger on "

• perf-loop

  Optimize a metric under its budget: benchmark the whole set under fixed conditions, fix the biggest hotspot, re-benchmark, until every measured path is under target with no functional regression. Use for latency, throughput, page-load, bundle-size, memory or test-speed targets. Trigger on "optimize

• repo-cleanup-loop

  Weekly repo hygiene: classify every branch, PR and worktree, recover valuable work first, then remove one class of stale items with evidence. HITL on anything ambiguous; never force-deletes unmerged work. Broader and recurring; one-off post-merge cleanup is git-post-merge-cleanup. Trigger on "weekly

• self-healing-loop

  Watch CI or production errors on a cadence; when an actionable error appears, trace the root cause, fix it in a branch, verify the signal passes, and open a PR. Never auto-deploys or auto-merges. Use for an unattended watcher over CI, logs or alerts. Trigger on "watch the logs and fix errors", "self

• madness

  The front door to the toolkit: reads what you want, picks the right starting skill (orchestrator, plan-builder, a loop, a role or workflow skill, or skill-explorer) and launches it. Confirms before anything expensive. Use when you don't know which skill to start with. Trigger on "/madness", "where d

• model-adaptation

  Adapt a prompt, skill, brief or agent scaffold when the model or host changes (new Claude version, DeepSeek, GPT, Gemini, Freebuff, unknown runtime): what to prune, what backfires, what long runs need. Owns model/effort tiering. Use for any adapt-for-a-cell request, even with no brief attached. Trig

• skill-catalog

  Keep the catalog in sync with disk: plugin.json's skills array and every advertised skill count are derived by `scripts/catalog.sh --sync`, never hand-edited. Use when adding, removing or renaming a skill, on mismatched counts or a CI "catalog drift" failure, or to ask how many skills there are. Tri

• skill-eval

  Measure whether a skill actually helps: build a `claude plugin eval` suite, run it with and without the skill under a required spend cap, and read the delta. Use when adding evals to a skill, converting evals.json, re-checking skills after a model upgrade, or deciding if a skill should be cut. Trigg

• skill-explorer

  Point the user to the right skill without running it: names the skill and says why it fits. Use when someone asks which skill handles a task, what skills exist, what a named skill does or how skills relate, or has forgotten a skill's name. Trigger on "which skill for this", "what skills do I have",

• skill-review

  Audit skills for quality, consistency and triggering: 'all' for an ecosystem scan (ownership conflicts, length outliers, weak triggers, dead references) or one skill for a deep review. Writes a markdown report plus JSON for skill-update. Use when asked to "audit skills", "review this skill", "which

• skill-triage

  Measure skill routing with TypeSafe's Jev decision model: which skill a request should load, accuracy against labeled routing cases, and description collisions where one skill steals another's triggers. Cheap and spend-capped. Use when skills misfire or under-trigger, or before and after editing des

• skill-update

  Apply changes to an existing skill: read a skill-review report or inline findings, draft edits with a recommended answer each, walk and apply them, then re-run lint. Use right after skill-review to ship its recommendations. Trigger on "apply the review", "update this skill", "fix the skill", "edit t

• skill-writer

  Write a new skill from scratch: SKILL.md with correct frontmatter, directory layout, file ownership, references and a triggering test. Use when creating a new role, meta, workflow or contract skill. Trigger on "write a new skill", "create a skill", "add a skill for X", "scaffold a skill". directorie

• orchestrator

  Coordinate multi-agent builds end to end: read the plan or MISSION.md, design contracts, dispatch role agents in parallel, gate on QA, ship. Sequential on hosts without subagents; uses the Workflow tool under ultracode or a workflow ask. Use for agent teams, parallel or swarm builds. Picks up after

• backend-agent

  Build API servers, business logic and data layers against authored contracts. Use in native orchestrator dispatch or explicitly requested solo backend implementation; load it when the orchestrator assigns this role or the user asks for it.

• code-review-agent

  Review a diff, PR or files on two separate axes: Standards (repo conventions and a code-smell baseline) and Spec (does it implement the issue or contract). Finds the change itself (working tree, staged diff or PR) when none is pasted. Read-only. Use for any code-review request. Trigger on "review th

• db-migration-agent

  Manage database schema migrations, seed data and schema evolution in native orchestrator dispatch or explicitly requested solo migration work. Preserve contract boundaries and approval before database mutations; load it when the orchestrator assigns this role or the user asks for it.

• docs-agent

  Generate project documentation, API docs, READMEs and changelogs in native orchestrator dispatch or explicitly requested solo documentation work. Read real contracts and code; load it when the orchestrator assigns this role or the user asks for it.

• frontend-agent

  Build user interfaces: pages, screens, components, routing, client state, responsive layout and page shells. Use for any frontend implementation task (React, Vue, Svelte, TSX, CSS) or an orchestrator dispatch to the frontend role. Runs the frontend source guards before reporting done. Trigger on "bu

• infrastructure-agent

  Build containerization, service orchestration, CI/CD and deployment configuration in native orchestrator dispatch or explicitly requested solo infrastructure work. Production actions require approval; load it when the orchestrator assigns this role or the user asks for it.

• observability-agent

  Set up logging, tracing, monitoring, metrics and alerting in native orchestrator dispatch or explicitly requested solo instrumentation work. Preserve business-logic ownership; load it when the orchestrator assigns this role or the user asks for it.

• performance-agent

  Design and execute performance tests, load tests and benchmarks in native orchestrator dispatch or explicitly requested solo measurement work. Freeze targets and obtain approval for nonlocal load; load it when the orchestrator assigns this role or the user asks for it.

• qe-agent

  Verify implementations match contracts, integrations connect and edge cases are handled in native orchestrator dispatch or explicitly requested solo QA. Own qa-report.json; disclose missing independence and blocked runtime checks. Load it when the orchestrator assigns this role or the user asks for

• security-agent

  Audit codebases for security vulnerabilities, auth defects and OWASP compliance in native orchestrator dispatch or explicitly requested solo read-only security review. Report findings, never fix production code; load it when the orchestrator assigns this role or the user asks for it.

• accessibility-audit

  Audit and fix web accessibility against WCAG 2.2 AA, Section 508 and the ADA: axe scan, keyboard, screen-reader and zoom checks, plain-language content and PDFs, source fixes, and an honest conformance report (ACR/VPAT draft). Use for any a11y review, fix or compliance question. Trigger on "accessib

• andon

  Run work through a live Andon board (the kanban that dispatches cards to Claude Code, Codex and Gemini on your machines): push a plan's cards with the skills each run must load, read the board, answer what needs you, administer a machine, and refresh this skill when Andon changes. Trigger on "/andon

• architecture-rescue

  Find deepening opportunities in a codebase (shallow modules, missing seams, leaky abstractions) and present them as numbered candidates for a grilling session before any interface is proposed. Use when the code feels tangled or hard to test. Trigger on "improve architecture", "rescue this codebase",

• artifact-publish

  Publish a visual or interactive result as a Claude Code Artifact: a self-contained HTML or Markdown page on a private, shareable claude.ai link instead of a local file. Use when a dashboard, report, chart, diagram or prototype reads better as a page. Not the claude.ai design canvas (claude-design-br

• caveman

  Ultra-compressed reply mode: cuts response tokens about 75% by dropping articles, filler and hedging while keeping code, errors and paths exact. Persists until "stop caveman" or "normal mode". Trigger on "caveman mode", "talk like caveman", "less tokens", "be terse", "/caveman". directories: [] patt

• class-extraction-guard

  Source-level gate for utility-class soup: the same long run of utility classes copy-pasted across files instead of extracted into a named class or component. Use when frontend work touches class markup, before calling UI work done, or as a wave gate. Sibling of design-token-guard (that checks which

• claude-design-brief

  Write paste-ready prompts for Claude Design (the claude.ai design canvas, not Claude Code) committed enough to skip its Q&A and start drawing. Use for hi-fi mockups, interactive prototypes, or safe/bold/experimental directions on the canvas. For Claude Code production builds use ui-brief. Trigger on

• cloudflare-pages-deploy

  Deploy static sites, previews and review hubs to Cloudflare Pages (free) with wrangler, then verify the public URL. This machine is already logged in with live projects, so check here before concluding Cloudflare is unavailable. Not Railway, here.now, or a private claude.ai page (artifact-publish).

• context-manager

  Manage context and session handoffs for long or multi-agent builds: pick a compaction strategy and write and validate a structured handoff file so a fresh session picks up cleanly. Use when the context is filling or work must continue in a new session. Trigger on "we're running out of context", "han

• dependency-coordinator

  Orchestrator-dispatched only. Authors the cross-package dependency manifest before parallel implementation agents are dispatched, preventing transitive version drift. Composed by orchestrator during multi-agent builds. Not user-invocable.

• deployment-checklist

  Run pre-deployment verification before pushing to staging or production — build, tests, env config, security, DB, infra, integration — and return a READY / NOT READY verdict. Trigger on: "pre-deploy check", "deployment checklist", "ready to ship", "is this ready for prod", "deploy readiness", "relea

• design-token-guard

  Source-level gate that stops inline styles and hardcoded colors from bypassing a project's design tokens. Framework-agnostic; auto-discovers the tokens. Use when frontend work touches colors or themes, before calling a UI task done, or as a wave gate. Render reviews can't see a hardcoded color. Trig

• diagnose-loop

  Disciplined bug diagnosis whose core is building a fast, deterministic pass/fail reproduction first; the later phases run on that signal. Use for hard bugs, flaky tests or performance regressions. Trigger on "diagnose this", "debug this", "why is this broken", "it sometimes fails", "this test is fla

• find-unknowns

  Turn unknown unknowns into known ones cheaply: a blindspot pass that finds the questions you don't know to ask in unfamiliar territory, and a comprehension quiz on a big diff before merging. Sibling of grill-me, which covers decisions you already hold opinions on. Trigger on "find my unknown unknown

• grill-me

  Interview the user relentlessly about a plan or design, one question at a time, each with a recommended answer checked against the code, until every branch of the decision is resolved. Use before non-trivial changes or when scope is fuzzy. Trigger on "grill me", "interview me about this", "challenge

• interactive-doc

  Produce documentation as an Obsidian-friendly markdown source plus a self-contained HTML companion built from it: render an existing research .md, or write both for a new doc. Use for architecture deep-dives, concept explainers with demos, or repo comparisons. Trigger on "interactive doc", "Obsidian

• living-plan

  Set up the living-plan convention: START-HERE.md, a roadmap, the REMAINING-WORK / COMPLETED-WORK / FUTURE ledger trio, and an intake loop so reports become tracked work. If this project lacks those files, set them up here rather than searching other repos. Trigger on "set up a living plan", "stop my

• llm-wiki

  Bootstrap and run an LLM-maintained knowledge base for a project or domain, a compounding wiki rather than RAG, with setup, ingest, query and lint modes. Trigger on "set up a wiki for X", "build me a knowledge base", "set up my second brain", and inside a wiki on "add this to the wiki" or "what does

• madness-skill-creator

  Create, edit and measure skills with Hermes Agent or Freebuff (Codebuff engine, any OpenAI-compatible endpoint): trigger evals with isolated snapshots and description optimization with strict train/dev/held-out split hygiene. Use to build or optimize a skill on those hosts. - Read - Write - Edit - G

• maintain-context

  Keep a project's CONTEXT.md glossary and docs/adr/ decision records current as understanding settles: add a glossary entry when a term resolves, an ADR only when its three-condition gate fires. Use after architectural discussions or when terms drift. Trigger on "update the glossary", "record this as

• mermaid-charts

  Create expert mermaid diagrams (flowchart, sequence, state, ER, Gantt, mindmap, block-beta), including dense 15-30+ node architecture maps that stay readable. Use whenever a system, process, architecture, data model or timeline should be visualized, even without the word "mermaid". Trigger on "diagr

• nano-banana

  Generate images with Google's Nano Banana (Gemini image generation) and save them to the project: product photos, hero banners, lifestyle shots, batches. Use for any Gemini image task. Trigger on "generate an image", "Nano Banana", "make a banner", "let's do images", "make me a picture of". director

• payload-cms

  Build, extend, deploy and debug sites on Payload CMS: collections, blocks, live preview, drafts, plugins, Postgres or Mongo, Vercel, migrations, and rebuilding an existing site onto Payload. Use when the user mentions Payload, payload.config.ts, @payloadcms packages, or a headless CMS for their site

• plan-builder

  Turn research docs, PRDs and conversational goals into structured project plans the orchestrator can execute. Use when there's source material plus a build request, or when orchestrator would be next but no plan exists. Produces the plan; orchestrator consumes it. Trigger on "make a plan", "plan thi

• plan-intake

  Turn a finished report (deep dive, audit, skill-review, QA findings) into approved entries in the project's living-plan ledger, in that ledger's own format. Use when a report should become tracked work instead of rotting. Trigger on "intake this report", "add findings to the plan", "turn this audit

• playwright

  Run browser E2E tests, capture screenshots and validate user flows with Playwright in visible Chrome. Use for end-to-end UI testing, visual review screenshots, responsive checks or accessibility audits, including qe-agent's browser testing. Trigger on "e2e test", "screenshot the UI", "click through

• project-profiler

  Analyze a codebase and generate a project profile (CLAUDE.md + profile.yaml) so agents have shared context. Use when onboarding a new project, generating CLAUDE.md, or profiling tech stack and conventions. Trigger on: "profile this project", "generate CLAUDE.md", "onboard this codebase", "describe t

• prose-slop-guard

  Gate that catches AI-sounding prose before it ships: em dashes, AI vocabulary (delve, leverage, seamless), "not X, it's Y", throat-clearing, filler, rule-of-three rhythm. A lexical checker plus a rewrite pass. Use on READMEs, docs, PR descriptions and posts before calling writing done. Trigger on "s

• railway-deploy

  Deploy projects to Railway — Dockerfile creation, railway.toml config, environment variables, multi-service setups (web + worker), and deployment via CLI or GraphQL API. Trigger on "deploy to Railway", "ship it on Railway", or "set up Railway hosting", and on questions about Railway configuration, h

• render-sanity

  Before-done visual sanity check for UIs that pass tests but break when clicked: stale mock IDs, lone ? or undefined where data belongs, dead links, Unauthorized shells. Apply it even without a browser or shell and report what you can't check as BLOCKED. Use when a build wraps up or tests pass but th

• repo-deep-dive

  Deep dive an open-source repo: combine a deep-research markdown with hands-on code analysis into a 12-14 document reference series, optionally compared against a reference project. Use to analyze, reverse-engineer or document how a project works. Trigger on "deep dive", "analyze this repo", "break d

• settings-consolidator

  Merge every .claude/settings.local.json under the home directory into the global one: dedupe permissions, collapse supersets, report by category. Also bootstraps permissions for unattended sessions from a 350+ command baseline with a deny list. Trigger on "consolidate settings", "merge permissions",

• setup-project-skills

  Bootstrap the per-repo config other skills read: domain doc layout (CONTEXT.md, ADRs), contract format and work-item tracker. Writes docs/agents/ plus an "Agent skills" block in CLAUDE.md or AGENTS.md without overwriting. Run once per repo. Trigger on "setup project skills", "configure project skill

• sync-skills

  Sync approved Skill-Madness skills into Claude Code and Cursor with links or resource-complete copies, honoring category or skill subsets and never touching unrelated skills without backed-up approval. Trigger on "sync skills", "link skills", "skill status", "unlink skills". directories: ["skills/wo

• ui-brief

  Write an opinionated, design-led brief for building or rebuilding a UI so it leads with the product's moat instead of a generic shadcn dashboard. Output is a standalone Markdown spec with a verifiable definition of done. Use before non-trivial UI work, greenfield or rebuild. Trigger on "write me a U

• use-freellmapi
• use-pxpipe
• website-walkthrough-video
• wiki-research
• work-item-brief
• yagni-gate
• zoom-out

Similar plugins