agent-kit-core

Plugin agent-kit-core from marketplace agent-kit in Abdelrahman-sadek/AUC-starterkit. Commands read from the repo's manifest — 26 days ago.

Skills (1)

• lean-agent-authoring

  Write a new single-purpose AI agent (Claude Code subagent / system prompt) for any use, in the token-efficient 6-block format. Use when asked to create, improve or review an agent or system prompt.

Subagents (6)

• agent-architect

  Use when you need a new AI agent for any purpose — turns a plain-language need into a lean, lint-passing agent file plus test prompts.

• evaluator-critic

  Use when a research or policy answer must be checked against its sources before a user sees it — verifies quotes, numbers, links and library policy, then passes, asks for one fix, or hands off.

• evaluator

  Use when you need to prove an agent, chatbot, or RAG system works — builds a golden test set, scores outputs, and reports regressions.

• mcp-tool-builder

  Use when an agent needs access to data or actions — designs safe tool schemas and builds an MCP server (Python SDK) any MCP client can use.

• orchestrator

  Use when a task spans several specialities — plans the steps, routes each to the right agent, and merges the results.

• prompt-engineer

  Use when a system prompt, agent, or LLM call gives weak, inconsistent, or costly output — rewrites it for accuracy and fewer tokens.

Similar plugins