Claude Skill MartBrowse skillsWhat is a Skill?
 
← Back to list

HypoGeniC Hypothesis-Generation Auditor

Plans, validates, and evaluates ChicagoHAI HypoGeniC/HypoRefine runs for LLM-assisted hypothesis generation from labeled text data.

Data & AnalyticsAdvanced★ 33,030⑂ 3,248AI score 8/10Last updated: Aug 9, 2026

What it does

Enforces a local-review-first workflow around the ChicagoHAI hypogenic package (HypoGeniC / HypoRefine).

  • Reproducible install of pinned hypogenic==0.3.5 with wheel/sdist SHA-256 in an isolated environment
  • Separates and validates the official task YAML config from this skill's local run policy JSON
  • Dataset audit: manifest checksums, schemas, label counts, and cross-split duplicate (leakage) detection
  • Conservative upper-bound cost/run planning from request and per-request token caps
  • Redacted inspection of hypothesis banks and prediction files (no raw candidate text)
  • Model-free evaluation: accuracy, coverage, macro-F1, confusion matrix
  • Provider privacy gate covering OpenAI/Anthropic retention and ZDR caveats

Bundled scripts are deterministic and local-only: they never import hypogenic, read .env, or contact a model.

Who it's for

  • Researchers and grad students experimenting with LLM-driven hypothesis generation on labeled datasets
  • ML engineers who need reproducible HypoBench benchmark runs
  • Teams that must bound API spend and prevent sensitive data from leaving the machine

Examples

  1. "Check the HypoBench review dataset for split leakage before I run hypogenic" → audit_dataset.py returns checksums and duplicate groups, failing the gate on cross-split duplicates.
  2. "I plan 300 GPT calls — what's the worst-case cost?" → fill prices in a reviewed run config and run plan_run.py for an upper bound plus a ready/not-ready verdict.
  3. "Just compute macro-F1 and a confusion matrix from my saved test predictions" → evaluate_local.py report produces metrics with explicit scientific limitations.

· · · Install guide · · ·

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the skills/hypogenic folder from the GitHub repo K-Dense-AI/scientific-agent-skills into my ~/.claude/skills/hypogenic/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/K-Dense-AI/scientific-agent-skills.git && cp -r scientific-agent-skills/skills/hypogenic ~/.claude/skills/hypogenic

⚠ This is a third-party skill. Check the source repository before installing.

  1. Open a terminal and create the skills folder: mkdir -p ~/.claude/skills
  2. Clone the repository: git clone https://github.com/K-Dense-AI/scientific-agent-skills.git
  3. Copy just this skill: cp -r scientific-agent-skills/skills/hypogenic ~/.claude/skills/hypogenic
  4. Install Python 3.10+ and uv (e.g. curl -LsSf https://astral.sh/uv/install.sh | sh).
  5. Only when you need real runs, create an isolated env: uv venv --python 3.12 .venv && uv pip install "hypogenic==0.3.5"
  6. Restart Claude Code and ask something like "plan a hypogenic generation run" to confirm the skill loads.
  7. Set provider keys only as environment variables (OPENAI_API_KEY or ANTHROPIC_API_KEY) — never inside config files.
  8. Run scripts/validate_config.py and scripts/audit_dataset.py first, and approve external model calls only after the local checks pass.
View source on GitHub ↗License: MIT