search-lossless-corpus

What problem does it solve?

This Skill solves the problem of getting inaccurate, generic answers from training data when asking about the Lossless Group's prior work, decisions, shipped features, and past issues. Without it, you would rely on stale or unverified information instead of the team's actual, citable internal records.

Core Features & Use Cases

• Multi-Collection Search: Queries four curated Chroma collections (context-vigilance-corpus, lossless-changelog, claude-code-sessions, claude-code-tool-traces) to cover decisions, shipping history, past sessions, and tool failures.
• Structured Search Loop: Uses a 4-step agentic workflow (decompose → execute → evaluate → synthesize) with a 5-query budget to find precise answers efficiently without wasting resources.
• Citation Discipline: Every claim is tied to a verifiable source path, timestamp, or repo slug, so you can confirm the answer in seconds.
• Use Case: If you ask "why did we choose PersistentClient for Chroma?", the skill pulls the exact decision from past Claude Code sessions, cites the source document, and lets you verify the context in seconds.

Quick Start

Use the search-lossless-corpus skill to find out why we chose PersistentClient for our local Chroma setup and when that decision was made.