メニュー
Curador do corpus RAG. Gerencia adição, organização e manutenção do conhecimento do projeto. Garante qualidade e acessibilidade.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | rag-curator |
| description | Curador do corpus RAG. Gerencia adição, organização e manutenção do conhecimento do projeto. Garante qualidade e acessibilidade. |
Skill responsável por curar e manter o corpus RAG do projeto.
corpus/nodes/decisions/corpus/nodes/learnings/This skill uses path_resolver.py to dynamically resolve project artifacts directory from settings.json:
# Resolve project directory
PROJECT_DIR=$(python3 .claude/lib/python/path_resolver.py --project-dir)
# ADRs are read from: ${PROJECT_DIR}/projects/{id}/decisions/
# Corpus is written to: ${PROJECT_DIR}/corpus/nodes/decisions/
Why this approach:
settings.json configuration (default: .project).agentic_sdlc/ (which is for framework files only)Note: Scripts automatically use path_resolver.py to find project directories.
# Index ADRs from project to corpus
# (automatically resolves paths from settings.json)
python3 .claude/skills/rag-curator/scripts/index_adrs.py --project-id PROJECT_ID
# Index all projects
python3 .claude/skills/rag-curator/scripts/index_adrs.py --all
# Validate corpus quality
python3 .claude/skills/rag-curator/scripts/validate_corpus.py
# Manual path resolution (if needed)
PROJECT_DIR=$(python3 .claude/lib/python/path_resolver.py --project-dir)
echo "Project artifacts: $PROJECT_DIR"
scripts/index_adrs.py: Index ADRs to corpusscripts/index_learnings.py: Index learnings to corpusscripts/validate_corpus.py: Validate corpus qualityOne-shot utility skill that migrates every agent under `.claude/agents/` to carry a tier-2 contract_version=1.0 frontmatter block, as required by Constitution Principle XI. Idempotent — agents that already have a contract are skipped verbatim. This skill is scheduled for removal after Phase B closes; its value is historical reproducibility.
Orchestrates GitHub Copilot Cloud Agent as the default strategy for parallel implementation work. Converts SDD tasks into GitHub issues, assigns them to @copilot, tracks agent jobs, reviews resulting PRs, and feeds back review comments as @copilot replies until merge-ready. This is the primary parallelization strategy for Phase 6 (Implementation) in Levels 2 and 3. Local parallel-workers (git worktrees) remains the fallback for tasks that require local environment access or credentials.
Integracao com BMAD Method para escala adaptativa e workflows guiados. Detecta nivel de complexidade e ajusta agentes automaticamente. Use quando: iniciar workflow, detectar complexidade, mapear agentes BMAD.
Avalia quality gates entre fases do SDLC. Verifica artefatos obrigatorios, criterios de qualidade, e aprovacoes necessarias antes de permitir transicao. Use quando: transicao entre fases, verificacao manual de gate, auditoria de qualidade.
Manages parallelization strategies (Agent Teams vs Parallel Workers) for SDLC phases. Detects optimal strategy based on phase, complexity, and task characteristics. Coordinates team creation, messaging, and task dependencies for Agent Teams. Falls back to parallel-workers for implementation phases requiring file isolation. Use when: coordinating parallel work, research phases, architecture discussions, quality reviews.
Analisa sessoes do Claude Code para extrair learnings e persistir conhecimento. Le arquivos de sessao em ~/.claude/projects/ e extrai decisoes, bloqueios e resolucoes. Invocado automaticamente pelo gate-check e orchestrator. Use quando: fim de fase, retrospectiva, analise de progresso.