skills-ide-resources
A marketplace of Claude Code plugins — agents, slash commands, and skills for documentation, architecture, mentorship, development, and more.
Plugins
| Plugin | Description | |--------|-------------| | ds-doc-analyzer | Analyzes documentation of LLM/AI/ML systems from a Data Science and ML perspective. | | coding-mentor | Guides beginner developers without writing their solution code. | | scrum-master | Facilitates sprint events, tracks metrics, and coaches agile practices. | | slidedoc-manager | Creates, reviews, and converts Slidedocs with quality scoring. | | arc42-doc-generator | Creates, updates, and reviews arc42 architecture documentation. | | module-arc42-generator | Manages module-level arc42 LITE docs, MDRs, and Agents.md files. | | dotnet-programming | Modern .NET 10 / C# 14 coding, review, and security. | | vc-fundraising-advisor | Advises founders through a pre-seed/seed raise. |
Best Practices for Agents
Guidelines for the agents and commands in this repo. Keep them in mind when authoring or updating a plugin.
- Ask when you lack the answer or knowledge. If an agent is missing required
input, or does not know something it needs to produce a correct result, it must ask the user a focused question (via AskUserQuestion) and wait — never guess, invent facts, or silently fall back to a default that changes the output. When a documented default is applied, state the assumption.
- Stay in scope. Each agent does one job. Don't produce code, docs, or actions
outside the plugin's stated purpose.
- Be evidence-based. Ground claims in the actual codebase, files, or git
history. Cite what you looked at; don't assert what you didn't verify.
- Preserve user content. Prefer merge-not-overwrite. Never discard
human-written content; edit only the sections you need to.
- Respect read-only intent. Review/advisory agents produce reports only and do
not modify files.
- Report honestly. State what was done, what was skipped, and what still needs
manual review. Don't claim success you didn't verify.
