grayskripko/marketing-skills

Free skills-only plugins in the Claude plugin directory. Each plugin lives in its own folder under plugins/ with its own README, license and privacy note; install them from the Claude plugin directory.
AI agents and skills
| Plugin | What it does |
|---|---|
| Agent Debugging | Debug a failed AI agent run from traces, logs, configuration, transcripts and output files; find the supported fault and get a repair or reproducible escalation. |
| Agent Design | Design an AI agent for one process: decide whether an agent is needed, then write its implementation specification. |
| Agent Evaluation | Plan agent evaluation and verify agent output against supplied requirements, artifacts, action receipts and escalation records. |
| Agent Guardrails | Define and check agent permissions, human approval and agent escalation for a real workflow. |
| Agent Recovery | Plan recovery of an interrupted AI agent run from checkpoints and action receipts; identify what can resume and which effects need checking before retry. |
| Instructions Cleanup | Fix conflicting CLAUDE.md and AGENTS.md instructions: map the active scopes, repair stale references and return a minimal patch that preserves approvals. |
| MCP Debugging | Troubleshoot MCP connection, authentication and missing-tool errors in my actual client from pasted traces, logs, configs, transcripts or files. |
| Plugin Packaging | Repair a skills-only Claude plugin that fails to install, load skills or find its resources. |
| RAG Evaluation | Evaluate and debug wrong RAG answers, retrieval quality, missing documents and grounded citations from the question, source documents, retrieved passages and response. |
| Tool Schemas | Fix AI agent tool contracts when connections work but tool choices, arguments or results fail. |
Author
Sergey Skripko. I build AI agents and fix poorly working skills.
If a plugin or a skill of yours works wrongly for your task, write in GitHub Discussions — I read every message. Bugs: GitHub Issues.