coding-agent-mcp-tools: Instructions file for Claude Code

coding-agent-mcp-tools — Claude Code Instructions

This repo is a curated list of coding agents and MCP tools. It has TWO special workflows: Idea Validation and Auto Research.

---

TOOL 1: Idea Validation (via idea-reality-mcp + README scan)

When someone says "validate idea: X" or "is X worth building?", run this flow:

Step 1 — idea_check via MCP

Call the idea_check MCP tool with the idea description. Use depth="deep" for thorough validation.

Step 2 — Scan THIS repo's list (THE MOST IMPORTANT STEP)

Go through the README.md of THIS repo (coding-agent-mcp-tools). For EVERY repo listed, do:

1. Fetch its GitHub URL (listed in README)
2. Read its README content (use WebFetch tool)
3. Check if the idea overlaps with what that repo does

Step 3 — Decision Logic

NEGATIVE (not worth building):

• If 2+ repos in our list already cover the idea's core capability
• State which repos cover it and why
• Format: "NEGATIVE — Already covered by [repo-name]: [reason from their README]"

POSITIVE (worth building):

• If NO repo in our list covers this specific angle
• Format: "POSITIVE — Gap identified: [reason]. Closest existing: [repo] but differs because [difference]"

PARTIAL (niche opportunity):

• Core idea exists but this specific angle is unique
• Format: "PARTIAL — Core exists in [repo], but [specific angle] is not covered"

Step 4 — Final Report Format

IDEA: [idea name]
DECISION: [POSITIVE / NEGATIVE / PARTIAL]

MCP Signal: [reality_signal]/100 — [duplicate_likelihood]
Top Similar (from MCP): [list top_similars]

Internal Repo Scan:
- [repo-name] ([link]): [relevant/not-relevant] — [one line reason]

REASONING:
[2-3 sentences explaining the decision]

VERDICT:
[Final recommendation]

---

TOOL 2: Auto Research (via autoresearch skill)

When someone says "research: X" or "find repos for: X", use the /autoresearch skill.

Research Pattern Examples:

RAG with vision support:

/autoresearch
Goal: Find repos in our list that support RAG with BOTH text AND vision/image documents
Scope: README.md (read-only, do not modify)
Metric: count of repos found matching criteria (higher is better)
Verify: node -e "const r=require('./research-results.json');console.log('SCORE: '+r.length)"