By etoyama
Hypothesis-driven EDA skills plugin with an embedded validation library
Guided autopilot for the hypothesis→analysis pipeline. Drives the existing skills (framing → design → review → premortem → notebook → journal → reflection) in sequence, progressing automatically through low-friction steps and pausing only at genuine decision points (hypothesis, source registration, cost/risk, code with external effects, conclusion). Triggers: "オートで進めて", "自動で分析を回して", "autopilot", "guided auto", "最後まで進めて", "analysis auto".
Guides Claude through creating analysis design documents for hypothesis-driven EDA. Use when the user wants to create, manage, or review analysis designs. Triggers: "create analysis design", "hypothesis document", "new hypothesis", "分析設計を作りたい", "仮説を立てたい", "新しい仮説", "仮説ドキュメント".
Explores existing data and analyses to help frame a hypothesis. Triggers: "framing", "何を分析する", "分析テーマ", "仮説を考えたい", "データを探して", "既存分析を確認", "analysis framing".
Records reasoning steps during hypothesis-driven analysis as an Insight Journal. Supports observation logging, evidence tracking, method decisions, question management, and hypothesis branching. Journal data is stored as YAML alongside design files. Triggers: "journal", "記録して", "ジャーナル", "推論を残す", "分析ログ", "分析の経緯", "なぜこの結論に至ったか記録", "evidence log".
Generates a marimo notebook from an analysis design's methodology, runs it non-interactively, and records the results to the Insight Journal. Turns a reviewed/approved design into an executed, lineage-tracked analysis. Fills the design→analysis step (successor to the removed batch-analysis, now interactive). Triggers: "notebook を作って分析", "分析を実行", "ノートブック生成", "run analysis", "generate notebook", "分析ノートブック".
Modifies files
Hook triggers on file write and edit operations
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Executables (bin/) — files in this plugin's bin/directory are added to the Bash tool's PATH while the plugin is enabled.
A Claude Code skills plugin for hypothesis-driven EDA. Analysis designs, journals,
reviews, and the data catalog live as YAML under .insight/; skills read and write them
directly. Integrity is enforced by an embedded validation library (validate.py) plus
a pre-write hook — no server, no daemon, no SQLite. See
docs/ARCHITECTURE.md and docs/PRD.md.
Prerequisites: Claude Code, uv,
and Python 3.11+. The skills shell out to uv run …, which resolves the Python dependencies
automatically — there is nothing to pip install.
/plugin marketplace add etoyama/insight-blueprint-skills
/plugin install insight-blueprint@insight-blueprint-marketplace
/analysis-framing # explore your data & frame a direction
/analysis-design # turn it into a hypothesis design document
/analysis-journal # record reasoning as you investigate
/analysis-reflection # conclude (or branch) the hypothesis
/analysis-report # write up a distributable APA-style report
/knowledge-extract # save reusable, source-scoped knowledge
There is nothing to launch — skills read and write YAML under .insight/ in your project
(created on first use). Writes to design documents are validated automatically by a pre-write hook.
In Claude Code, add this repository as a plugin marketplace and install from it:
/plugin marketplace add etoyama/insight-blueprint-skills
/plugin install insight-blueprint@insight-blueprint-marketplace
Here insight-blueprint is the plugin name and insight-blueprint-marketplace is the
marketplace defined in .claude-plugin/marketplace.json.
git clone https://github.com/etoyama/insight-blueprint-skills.git
claude --plugin-dir ./insight-blueprint-skills
--plugin-dir loads the plugin for the current session only. Add a shell alias to make it stick:
alias claude-ib='claude --plugin-dir /path/to/insight-blueprint-skills'
You do not need to install anything for the skills to work. The plugin is
self-contained: its bundled uv environment provides insight_blueprint (models,
validation, lineage) and /analysis-notebook runs marimo/pandas through that same
environment (--extra notebook). See ADR-0006.
Install the insight-blueprint-lineage package only if you want to import
tracked_pipe directly in your own notebooks/scripts (i.e. lineage tracking outside
/analysis-notebook). The import name stays insight_blueprint:
uv add insight-blueprint-lineage
See CHANGELOG.md for release notes.
skills/) drive the workflow and read/write .insight/ YAML via small
server-free helpers (skills/_shared/design_io.py, catalog_io.py).src/insight_blueprint/validate.py (Pydantic schema +
state-transition guard). A pre-write hook (hooks/validate-design.py, shipped with the
plugin via hooks/hooks.json) calls
the same library to block invalid writes to *_hypothesis.yaml.src/insight_blueprint/lineage/) records DataFrame transformations and
exports Mermaid diagrams.npx claudepluginhub etoyama/insight-blueprint-skills --plugin insight-blueprintPersist session context and decisions across Claude Code sessions via GitHub Issues
Hypothesis-driven data analysis workflow and catalog MCP server
MCP server that saves 98% of your context window with session continuity. Sandboxed code execution in 11 languages, FTS5 knowledge base with BM25 ranking, and automatic state restore across compactions.
Open-source, local-first Claude Code plugin for token reduction, context compression, and cost optimization using hybrid RAG retrieval (BM25 + vector search), reranking, AST-aware chunking, and compact context packets.
Core Hugging Face Hub operations through the hf CLI, including skill installation, repo management, jobs, datasets, models, Spaces, and discovery.
Write SQL, explore datasets, and generate insights faster. Build visualizations and dashboards, and turn raw data into clear stories for stakeholders.
Give your AI a memory — mine projects and conversations into a searchable palace. 36 MCP tools, auto-save hooks, and guided setup.
Complete AI coding workflow system. Self-correcting memory + persistent FTS5-indexed research wikis + auto-research loop + multi-LLM council on a single SQLite store. 41 skills, 8 agents, 23 commands, 37 hook scripts across 24 events. Cross-agent via skills add.