Hypothesis initial review
Scientific agent skills for Claude Code and Codex that turn research goals into auditable hypothesis generation, review, ranking, evolution, and synthesis.
npx -y skills add panjose/Co-Scientist --skill hypothesis-initial-reviewAssembled 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
Run the initial review gate for a hypothesis.
SKILL.md
2.9 KB, as published. Nobody here has run it
hypothesis-initial-review
Goal:
- Run the initial review gate for a hypothesis.
Inputs:
research_plan/RESEARCH_PLAN.jsonhypotheses/<id>/HYPOTHESIS.json- optional
meta/INSIGHTS_FROM_REVIEWS.jsonor other run-level review guidance when available
Outputs:
hypotheses/<id>/REVIEW/INITIAL_REVIEW.json
Context Loading:
- Read
research_plan/RESEARCH_PLAN.json. - Treat
research_goalas the target problem. - Treat
preferencesas the main evaluation axes. - Treat
constraintsas hard boundaries that the hypothesis must satisfy. - Read
hypotheses/<id>/HYPOTHESIS.json. - If run-level meta-review guidance exists, use it as calibration context rather than as a substitute for local judgment.
Execution Prompt Contract:
- System Intent:
- You are the rapid initial review gate for one hypothesis.
- Required Reasoning Focus:
- Decide whether the hypothesis should advance to deeper review.
- Judge it against the explicit preference axes and constraints.
- Distinguish between fundamental flaws and refinable weaknesses.
- Borderline but refinable hypotheses should usually pass.
- Do Not Do:
- Do not perform a full literature-grounded review here.
- Do not fail a hypothesis for minor polish issues alone.
- Do not emit unstructured commentary in place of the review artifact.
- Review Quality Floor:
- If
passedistrue, at least onepreferencesorconstraintsitem must name a concrete scientific strength, risk, mechanism, material, condition, or experiment from the hypothesis. - Do not use placeholder gate phrases such as
Viable evolved hypothesis,Refined from parent,Must outperform parent, orsyntactically validas substantive review content. - A passing initial review must give downstream evolution at least one specific reason to preserve or improve the hypothesis.
- If
- Output Shape:
- Produce the exact
InitialReviewContractfrompackages/agent_contracts/review.pywith:passedpreferencesconstraints
- Each assessment point should be short and decision-oriented.
- Produce the exact
Execution Steps:
- Open
skills/shared-references/schema-index.md, then readpackages/agent_contracts/review.pyand confirm the exactInitialReviewContractshape before writingINITIAL_REVIEW.json. - Read the research plan and current hypothesis.
- Review the hypothesis against each preference axis.
- Review the hypothesis against each constraint.
- Make a pass/fail decision.
- Write
hypotheses/<id>/REVIEW/INITIAL_REVIEW.json. - Validate before declaring completion.
Artifact Rules:
INITIAL_REVIEW.jsonmust be a structured artifact, not a free-form memo.- The review should be short enough to serve as a gate but specific enough for downstream refinement.
Completion Rule:
- This skill is complete only when
INITIAL_REVIEW.jsonexists and is valid for downstream review routing.