mind-wasp-drone — Claude Code subagent: cognitive-layer review
mind-wasp-drone — a subagent for auditing and tuning everything LLM in your codebase: agent routing, prompt cascades, RAG, memory tiers, tracing, and streaming responses. Call it when you work on AI code, retrieval, or memory.
AGPL-3.0-or-later ★ 96 updated 2026-10-18
/plugin marketplace add legioncodeinc/vibe-coding-tools
/plugin install data-and-ai-tools@wasp-nestWhat mind-wasp-drone does
mind-wasp-drone owns the cognitive layer: coach-agent lineups, five-tier prompt cascades, three-tier memory (ephemeral cache, Postgres, retrieval substrate), per-call tracing of LLM invocations, retrieval and routing evaluation, multimodal pipelines, and orchestration. It checks each subsystem as a stack-neutral architecture first, then against your project's actual choice.
This project uses Neon Postgres plus pgvector by default, via vector-store-wasp-drone and retrieval-wasp-drone. If you already run Qdrant, Cohere, Valkey, or OpenRouter, that alternative stack remains fully documented and valid. The subagent stays out of chat UI, table indexing, security audits, and feature specs — other agents handle those.
Invoke for RAG audits, memory refactors, GraphRAG enablement, prompt cascade tweaks, streaming LLM responses over SvelteKit +server.ts on Vercel, or whenever you touch the cognitive layer in a PR.
What it helps with
- Audit a RAG system and verify two-stage recall-then-rerank retrieval
- Design a coach-agent architecture with per-call routing traces and metrics
- Build a three-tier memory system across LLM invocations
- Enable GraphRAG or swap retrieval substrates without data loss
- Stream an LLM response to a client over SvelteKit on Vercel
Example prompts for the agent
- Audit the RAG pipeline with a query and check recall-rerank stages
- Review the prompt cascade and agent routing traces
- Enable GraphRAG while preserving the current Postgres retrieval
How to install mind-wasp-drone
1Plugin marketplace
/plugin marketplace add legioncodeinc/vibe-coding-tools/plugin install data-and-ai-tools@wasp-nest2Manually: put the files in place
.claude/agents/mind-wasp-drone.mdFor all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention
Manually: put the files in place
.codex/agents/mind-wasp-drone.tomlCodex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/
Manually: put the files in place
.cursor/agents/mind-wasp-drone.mdCursor also reads .claude/agents/
Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean
Questions about mind-wasp-drone
How do I install mind-wasp-drone in Claude Code?
Run /plugin marketplace add legioncodeinc/vibe-coding-tools, then /plugin install data-and-ai-tools@wasp-nest and add .claude/agents/mind-wasp-drone.md to your config. For Codex, use .codex/agents/mind-wasp-drone.toml.
What does mind-wasp-drone review?
The cognitive layer: agent routing, prompts, RAG, memory tiers, LLM-call tracing, and response streaming. It examines architecture as a concept first, then checks your project's specific stack.
Which retrieval substrate does it use by default?
Neon Postgres with pgvector, implemented via vector-store-wasp-drone and retrieval-wasp-drone. Qdrant, Cohere, Valkey, and OpenRouter are fully supported alternatives for projects already running them.