Menu
SkillsMP has collected 94 skills from wcygan/agent-skills. Open a skill to review its source and details.
Showing 40 of 94 collected skills.
Use PromptKit’s official docs, machine-readable indexes, registry items, blocks, primitives, and source code to answer, debug, or implement React AI UI requests involving prompt-kit or prompt-kit.com. Trigger when a request names PromptKit, a PromptKit…
Coordinate a bounded multi-agent software task from decomposition through verified integration. Use for concurrent workers, dependent tasks, or shared integration ownership.
Create a beginner-friendly visual explainer with big pictures and few words as one self-contained HTML document. Use for ELI5 explanations of modules, tradeoffs, incidents, and unfamiliar topics.
Rewrite supplied text or Markdown in plain language while preserving facts, technical artifacts, structure, and the source language. Use when a user asks for a simpler explanation, document, or agent response without a meaning change.
Audit and design observability for one request, event, job, or state change across an application or distributed system using logs, metrics, traces, correlation identifiers, audit records, and operational artifacts. Use when making a flow measurable, defining…
Inspect the current GitHub pull request or stacked pull request as a read-only terminal dashboard, combining review threads, comments, CI, stack health, merge order, naming, visual proof, description quality, testing evidence, and split signals. Use when…
Assess and explain local and remote gh-stack health, topology, trunk freshness, atomic merge readiness, synchronization, rebase needs, recovery, CI shape, and safe next commands. Use when a stacked pull request may be stale, divergent, conflicted, misordered,…
Scaffold and compose a router-style skill in .agents/skills — validate the name, generate the SKILL.md plus references/ and scripts/ tree with a UV-runnable Python script, and write the body to the writing-for-agents discipline. Use when creating a skill…
Improve an agent harness through bounded experiments over a batch of traces or evaluations: cluster failures, classify layers, test one harness-variable hypothesis, cold-replay candidates, and require held-out or negative regression evidence. Use when…
Capture browser proof with Playwright CLI, prepare and upload images or videos for GitHub pull requests, then review or update PR descriptions. Use when a PR needs screenshots, recordings, media conversion, attachment recovery, before/after evidence, or…
Automate browser interactions, test web pages and work with Playwright tests.
Evaluate one pull request for concise description structure, visual proof, testing evidence, naming consistency, reviewability, code-quality signals, CI state, stack context, and split signals. Use when preparing, checking, or improving an individual PR,…
Audit a workflow graph for reachability, guards, terminal paths, cycles, retries, state, fanout, agent routes, authority, observability, and proof. Use when a graph can stall, duplicate work, use a wrong child route, lose state, evade policy, or lack…
Choose a control model for one bounded workflow: deterministic code, DAG, cyclic state graph, dynamic graph, simple agent loop, or multi-agent workflow. Use when orchestration, retries, state, approvals, agent discretion, fanout, recovery, or coordination…
Design a regression-oriented evaluation suite for an LLM, RAG pipeline, tool-using agent, or multi-agent workflow by defining behavioral cases, datasets, deterministic and model-graded oracles, trace or receipt requirements, stochastic thresholds, framework…
Design one bounded graph-shaped workflow as a read-only Graph Design Pack covering state ownership, node and typed edge contracts, loops, fanout, recovery, authority gates, observability, and verification. Use after a graph control model is justified, or when…
Evaluate and diagnose one agent workflow across instructions, context, route selection, tools, orchestration, durable events, and user output. Use for surprising results, wrong child routes, tool misuse, model migrations, or state disagreements. Produce an…
Ideate and structure higher-order agent skills that compose existing skills as orchestrators, routers, or decision frameworks. Use when designing a reusable workflow skill from several specialist skills, deciding whether composition is warranted, defining…
Model concurrent or distributed behavior with actors, state, operations, invariants, causal order, and counterexample schedules. Use for races, lost updates, deadlocks, duplicates, ordering, cancellation, retries, fanout, fanin, or nested agents. Separate…
Select and verify one agent model route. Cover provider, model, reasoning effort, service tier, role, context fork, tools, permissions, and fallback. Use before heterogeneous dispatch, route changes, route comparisons, inheritance diagnosis, unsupported-route…
Simplify one bounded code target and its same-behavior tests through verified edits. Use it when the user asks to simplify code, remove dead code or unnecessary indirection, flatten control flow, or consolidate repeated meaning. Also use it for bounded…
Audit a local project skill catalog for overlapping, fragmented, or poorly routed skills. Use when a workspace has folders such as `.agents/skills`, `.claude/skills`, `.codex/skills`, or `skills/`, users cannot find the right skill, or maintainers need safe…
Distill long responses or supplied text into a concise Human Brief with the gist, key points, practical meaning, actions, and important caveats. Use when the user asks for a summary, TLDR, salient points, plain-language explanation, executive brief, or…
Mine bounded local agent transcripts for recurring failures, verified workarounds, and reusable workflow knowledge. Use when reviewing Codex rollout sessions or other supported transcripts, including locating tasks by working directory, date, thread…
Explain technical topics with concise, copyable visuals. Use when structure, flow, state, ownership, comparison, or change is easier to understand visually.
Ask which skill or flow fits your situation. A router over the skills in this repo.
Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.
Implement a piece of work based on a spec or set of tickets.
Plan a huge chunk of work — more than one agent session can hold — as a shared map of decision tickets on your issue tracker, and resolve them one at a time until the way to the destination is clear.
Generate an interactive bash wizard that walks a human through steps only they can perform. Use when provisioning infrastructure, setting up credentials or CI secrets, walking an unfamiliar third-party dashboard, or running a one-off migration or cutover.…
Identify and rank hill-climbing opportunities by connecting desired outcomes to numeric target, driver, and guard metrics, measurement methods, and editable levers. Use when a codebase, system, deployment pipeline, or developer experience should improve, but…
Explore a broad topic and produce a landscape map of core ideas, vocabulary, viewpoints, open questions, and representative resources. Use for broad orientation, learning paths, or discovery before a bounded research question exists.
Inspect and improve an application's development loops from a change to trustworthy feedback. Use when setup, startup, fixtures, tests, builds, browser actions, logs, validation, or cleanup consume repeated developer or agent effort. Add the smallest useful…
Design, inspect, debug, and maintain justfile command workflows. Use when adding or changing just recipes, parameters, dependencies, modules, environment loading, shells, working directories, or project command entrypoints.
Guide Python work with uv across Python versions, PEP 723 scripts, uvx and uv tool commands, pyproject.toml projects, dependencies, uv.lock, configuration, indexes, resolution, and caching. Use when choosing or applying a uv workflow, maintaining uv-managed…
Deploy and live-test one bounded application change in an isolated local environment. Exercise relevant public paths and feature-flag states. Produce durable, reproducible evidence with exact Bash commands, requests, responses, observed outcomes, cleanup, and…
Audit code, services, workflows, and system designs for fragile states, failure amplification, weak recovery, poor debuggability, and difficult verification. Use before implementation, launch, migration, or reliability work when a system may be easy to break…
Run an explicitly requested, bounded software delivery loop from deliverable framing through current-state inspection, vertical implementation slices, and evidence-backed verification. Use for one scoped local change that should progress autonomously,…
This skill encodes Emil Kowalski's philosophy on UI polish, component design, animation decisions, and the invisible details that make software feel great.
Manages shadcn components and projects — adding, searching, fixing, debugging, styling, and composing UI, including chat interfaces. Provides project context, component docs, and usage examples. Applies when working with shadcn/ui, component registries,…