autosnippet-candidates

What problem does it solve?

AutoSnippet candidates are the structured, review-ready inputs that turn scattered code observations into high-quality Recipes. This Skill bridges the gap between raw code analysis and knowledge-base contributions by producing multi-layer metadata, scoring, and context links that reduce duplicates and speed up governance.

Core Features & Use Cases

• Layered information extraction (Layer 1-3) from code and docs to produce comprehensive Candidate payloads.
• Multi-dimensional quality scoring (code quality, documentation quality, project adaptability) to prioritize reviews.
• Single-file and batch Target scan workflows to scale candidate generation across projects.
• Contextual similarity checks against existing Recipes to detect duplicates, conflicts, or opportunities for merging.
• MCP tool integration for candidate submission to a dashboard with envelope metadata.
• Draft and annotations support for human-in-the-loop review.

Quick Start

• Run a single-file scan on a target file such as Sources/Network/RequestManager.swift to generate candidates.
• Or perform a batch Target scan by selecting a target with autosnippet_get_targets, obtaining files via autosnippet_get_target_files, then submitting candidates with autosnippet_submit_candidates.
• Review the Candidates in the Dashboard and approve or adjust before publication to the knowledge base.