github-research_skill

Overview

This skill coordinates external GitHub and library documentation research before development when the AI model lacks sufficient knowledge. It enforces a strict decision rule about when external research is allowed and which tools to try in order. The goal is to produce concise, machine-readable research artifacts that the cc10x router and agents can consume.

How this skill works

The skill first verifies an actual knowledge gap or an explicit user request, then probes Octocode MCP to confirm availability. It runs a tiered fall-back sequence (Octocode MCP → Context7 MCP → WebFetch) and captures focused findings, gotchas, and minimal code snippets. Finally, it saves the research into docs/research/, updates activeContext, and extracts reusable patterns into patterns.md before handing results to the router.

When to use it

• User explicitly asks for external research ("research X", "find on GitHub", "how do others")
• Technology or library released after 2024 or outside the AI cutoff
• Complex integrations with real uncertainty (auth, payments, real-time systems)
• Local debugging failed repeatedly with external service errors
• When the planner or bug-investigator requires external evidence or examples

Best practices

• Always confirm an AI knowledge gap before starting external research
• Try Octocode MCP first and verify availability with a simple package lookup
• Fall back to Context7 MCP, then WebFetch only if prior tiers fail
• Keep research goals explicit: mainResearchGoal, researchGoal, reasoning
• Save research files under docs/research/YYYY-MM-DD-<topic>-research.md immediately
• Atomically update activeContext.md and patterns.md in the same execution block after saving

Example use cases

• Research integration patterns for a new post-2024 API and save a minimal summary for the planner
• Investigate recurring external error messages from a third-party service after local debugging fails
• Collect best-practice snippets and gotchas for a complex OAuth or payment flow before implementation
• Find real-world examples on GitHub for an unfamiliar framework and extract applicable patterns
• Assemble concise references for the bug-investigator to diagnose external-service failures