Release v1.0.0 — Knowledge Base Management · bcbeidel/dewey

Complete rewrite of Dewey from a context-optimization tool (analyze/split) into a
knowledge-base management plugin for Claude Code.

Added

Curate skill — single entry point for all knowledge-base content operations:
discover domains, scaffold structure, add/update topics, ingest URLs, manage
proposals and curation plans
Health skill — three-tier validation and reporting system:
- Tier 1: 18 per-file validators and 6 cross-file validators (deterministic, CI-friendly)
- Tier 2: LLM-assisted pre-screener with 9 triggers for nuanced quality checks
- Tier 3: Human decision queue for subjective judgment calls
Report-issue skill — submit bug reports, feature ideas, or feedback to GitHub
Auto-fix — automated remediation for common Tier 1 issues
History tracking — snapshot-based health history for trend analysis
Utilization tracking — PostToolUse hook logs knowledge-base reads; feeds
recommendations for stale-high-use, expand-depth, low-utilization, and
never-referenced files
Readability checks — Flesch-Kincaid grade-level bounds per content depth
Source quality validators — placeholder detection, source diversity,
citation grounding, and source accessibility checking
Duplicate content detection — exact paragraph hashing plus Jaccard similarity
Naming convention enforcement — directory and file slug validation
Curation plan workflow — structured planning before content creation
Proposal lifecycle — propose, review, and promote content through validation gates
12 design principles grounded in agent context research and cognitive science

Consolidated explore, init, and curate skills into a single curate skill with
free-text intake and intent classification
Renamed all "KB"/"kb" references to "knowledge base"/"knowledge-base" throughout
the codebase (identifiers, CLI flags, marker comments, file names, prose)
Migrated all SKILL.md files to pure XML structure