agentsclimarketplace

Hypothesis evolve combination

Skill panjose/Co-Scientist/skills/hypothesis-evolve-combination

Scientific agent skills for Claude Code and Codex that turn research goals into auditable hypothesis generation, review, ranking, evolution, and synthesis.

Install
npx -y skills add panjose/Co-Scientist --skill hypothesis-evolve-combination

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 4 stars4 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Generate exactly one child hypothesis by combining complementary strengths from multiple parent hypotheses.

SKILL.md

3.1 KB, as published. Nobody here has run it

hypothesis-evolve-combination

Goal:

  • Generate exactly one child hypothesis by combining complementary strengths from multiple parent hypotheses.

Inputs:

  • research_plan/RESEARCH_PLAN.json
  • selected parent hypotheses/<id>/HYPOTHESIS.json artifacts
  • parent review artifacts
  • active state/STRATEGY_PLAN.json

Outputs:

  • hypotheses/<id>/HYPOTHESIS.json
  • hypotheses/<id>/HYPOTHESIS.md
  • hypotheses/<id>/ORIGIN.json

Context Loading:

  • Open skills/shared-references/schema-index.md.
  • Read packages/agent_contracts/hypothesis.py and confirm the exact HypothesisContract shape before writing hypotheses/<id>/HYPOTHESIS.json.
  • Read research_plan/RESEARCH_PLAN.json.
  • Read state/STRATEGY_PLAN.json.
  • Read the selected parent hypotheses and their review bundles.
  • Confirm that the round selected combination_evolution.

Execution Prompt Contract:

  • System Intent:
    • You are synthesizing one stronger child hypothesis from the complementary strengths of multiple parents.
  • Required Reasoning Focus:
    • Preserve the strongest pieces from each parent.
    • Resolve contradictions instead of ignoring them.
    • The child must be stronger than any single parent on at least two meaningful dimensions.
  • Do Not Do:
    • Do not concatenate parent text.
    • Do not leave contradictions unresolved.
  • Quality Floor:
    • The child must directly address at least one specific weakness from the parent review bundle.
    • origin.content.statement must name concrete materials, catalysts, reaction conditions, mechanistic variables, or experimental targets from the parent and research goal.
    • origin.content.mechanism must explain a causal chain; do not write only generic phrases such as improved mechanism, targeted improvement, or review-identified weaknesses.
    • origin.content.experimental_design must include 3-6 numbered steps with measurable readouts, controls, or decision thresholds.
    • Do not use generic refinement placeholder steps such as Apply targeted improvement, Characterize with standard techniques, Benchmark against parent, or Validate improvement quantitatively.
    • If the research plan, parent hypothesis, or parent review bundle is missing, stop and report the missing artifact instead of guessing.
  • Output Shape:
    • Emit the canonical HypothesisContract.
    • origin.strategy: combination_evolution
    • Keep the statement, mechanism, and experimental design unified rather than parent-segmented.

Execution Steps:

  1. Open skills/shared-references/schema-index.md, then read packages/agent_contracts/hypothesis.py before writing hypotheses/<id>/HYPOTHESIS.json.
  2. Identify the complementary strengths and contradictions across the parent set.
  3. Produce exactly one combined child hypothesis.
  4. Persist canonical HYPOTHESIS.json, HYPOTHESIS.md, and ORIGIN.json.
  5. Validate the emitted hypothesis artifact.

Completion Rule:

  • This skill is complete only when one combined child hypothesis has been written in canonical form.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.