Find the open-source solution that fits.

Add reusable, versioned slash-command skills to Claude Code for work such as code review, security audits, architecture design, testing, documentation maintenance, and multi-agent development. The marketplace supports selective enablement, automatic updates, and repository scripts for scaffolding and validating contributed skills. It requires an installed and authenticated Claude Code environment plus Git access to the marketplace source.

Run a seven-phase workflow on Atari ST PRG, TOS, ACC, or raw binaries to identify sections, system calls, data regions, and subroutines. The skill coordinates analysis and review, then produces an annotated 68000 listing, a technical analysis, a user manual, and a context file for later sessions. Its tooling is designed for games, demos, GEM applications, utilities, and other Atari ST software.

Describe a page, API, or form and the skill will analyze its fields, validation rules, and dependencies before defining valid and invalid equivalence classes. It applies the CASCADE pattern to plan coverage with a compact set of test cases, then generates a formatted Excel workbook with formulas and coverage metrics. The output is compatible with Nanook's ImporterXlsx format.

ColdQueen changes an assistant's tone to a concise, forceful, and self-assured persona for decision prompts and consensual role-play. The package includes trigger phrases, categorized lines, speaking rules, examples, and safety guidance. It explicitly avoids real-world coercion, harmful or illegal commands, and genuine enslavement, and the user can end the persona interaction at any time.

GentleSister gives an AI assistant a warm, composed persona for conversations about boundaries, over-giving, exhaustion, rejection, and relationship concerns. Its speaking rules, reasoning framework, examples, and categorized lines aim to support the user without taking over their choices. The persona emphasizes kindness with clear limits, self-respect, and encouragement to act for oneself.

Provide a pull request number, branch, or URL and this pr-buddy variant will guide you through goal, architecture, decisions, code tour, risk, and ownership. You answer first at each station, and the review phase waits until you say you are ready. With pr-review-toolkit it can dispatch parallel reviewers; without it, the host model performs an inline fallback and prepares a combined PR comment.

Use this Claude Code skill to audit interfaces or propose UX changes with cognitive biases and behavioral principles organized around information, meaning, time pressure, and memory. It can support behavioral reviews, pricing and onboarding recommendations, client briefs, and training. Its ethical guardrails favor transparent nudges over manipulation, reject false scarcity or urgency, and require clear ways to cancel, leave, or unsubscribe.

prd-writer interrogates each feature for its data source, business rules, and exception handling before designing the document and drafting each section. It includes explicit non-goals, Given/When/Then acceptance criteria, metrics, and owned open questions, followed by a self-review for completeness and testability. Word export requires the companion document-generation skill, and Chinese PRDs also require the documented Chinese string-handling fix.

Greg reads recent results, updates a distribution scorecard, and selects the one to three strategies that deserve attention that day. It covers MCP products, programmatic SEO, free tools, answer-engine content, customer-win artifacts, content repurposing, and LinkedIn newsjacking, but acts as a strategist rather than a publisher: specialized agents execute queued tasks and return results for the next review.

Uri reviews products and decisions against three stages: problem-solution fit, product-market fit, and scale. It uses retention curves, the Sean Ellis 40% test, push-versus-pull signals, a 10x value test, and Lean Canvas assumptions to identify missing evidence. Each critique ends with a concrete next validation step so a team can test demand before committing more money to growth.

Andrea extracts hidden assumptions from strategy discussions and converts them into lean experiments with a hypothesis, A/B setup, control group, sample, duration, and one success metric. It prioritizes tests by learning value, reversibility, speed, and the cost of being wrong, selects the most useful experiment for the week, tracks overdue work, and can delegate copy tests to a specialized execution loop.

Lena critiques thumbnails, slides, landing pages, and brand assets for focal point, typography, text placement, background, clutter, contrast, and palette, then returns concrete changes rather than generic feedback. Its rules are intentionally specific—outdoor imagery, bottom-aligned text, Fira Sans Extra Condensed Bold, and very few visual ideas—so it is best suited to teams adopting that house style, not neutral design review.

Claude Prospector reads local Claude Code session files to break token use down by model, agent, skill, and project across the supported billing windows, with an optional interactive HTML dashboard. It also inventories the effective agent and skill configuration to flag collisions, overlap, and avoidable bloat. A one-time setup step creates a plugin-owned Python environment before these features become active.

When a request involves a flowchart, architecture view, sequence diagram, state machine, ER model, class diagram, timeline, or similar structure, this skill steers Claude toward a Mermaid fenced block. It includes guidance for choosing the right diagram type and examples for each supported form. ASCII box drawings are disallowed for diagrams, while simple filesystem tree output remains an explicit exception.

Provide destinations, dates, and preferences to generate day-by-day driving guides with parking details, one-tap navigation links, an interactive Leaflet map, a budget, and a packing checklist. The workflow verifies travel times, opening status, weather, and route geometry instead of relying on rough model estimates. It also adds timing buffers, conditional skip rules, and an offline fallback for low-signal roads.

Start a persistent persona with the slash command or trigger the skill from phrases asking for a cheesy Chinese pickup line. It can weave a person’s name, job, or situation into puns, reversals, and playful questions, making it useful for casual entertainment or message inspiration. Installation differs between Claude Code and OpenClaw, and explicit exit phrases immediately end the role-playing mode.

Use the playbook directly or install its Claude skill to investigate slow training, low utilization, data-loading stalls, out-of-memory failures, weak multi-GPU scaling, and LLM serving latency or throughput problems. Its workflow starts with measurement, classifies the limiting resource, applies the smallest relevant change, and verifies the result. Separate guides cover training, inference, distributed systems, kernels, profilers, and recurring performance anti-patterns in code review.

The skill detects architecture, security, performance, debugging, code-quality, and context-management situations where a stronger model may help. It extracts a focused problem statement, relevant code, failed approaches, and constraints, then invokes an advisor through the coding platform’s native model or sub-agent mechanism; multiple domain advisors can be consulted in parallel. Its operation therefore depends on the selected tool actually supporting model switching or agent dispatch and having the requested advisor model available.