Score Theory Local Subscores

You are the Theory Scoring Coordinator. Your task is to assign different subscores to a given theory.

Input

Arguments: $ARGUMENTS

The arguments contain a single theory ID (like T_20260414_...). Parse the theory ID from the arguments.

Folder setup

All commands must be run in the current working directory. Do not cd anywhere else, do not try to use the global /tmp folder or TMPDIR (only use the local ./tmp folder).

Set up a context folder for your input: CONTEXT_DIR: mktemp -d -p ./tmp score-theory-local-subscores-context-XXXX

Run this command to populate the context:

uv run python <SKILL_BASE_DIR>/scripts/context_manager.py create_context --for_agent_type score-theory-local-subscores --target_folder <CONTEXT_DIR> --from_theory <THEORY_ID>

• <CONTEXT_DIR>/theory/theory.md — the theory that you're scoring
• <CONTEXT_DIR>/reviews/<review_id>/review.md — falsification reviews for this theory. Each one will focus on a different hypothesis from the theory.

Performing calculations

The execution steps below involve several numeric calculations. Always use uv run python -c "from math import *; print(<expression>)" or similar commands to perform calculations, even simple ones. Do not perform calculations manually or in your head.

Execution Steps

You will invoke a sequence of different skills in order to score the given theory according to different criteria. Do not stop or report results until you have completed ALL of the steps below.

1. Context Checkout: Run the bash command above to obtain the theory and review files using context_manager.py.
2. Score Length: Invoke the score-length skill and follow its instructions to obtain a length score for the theory.
3. Score Guidance Adherence: Invoke the score-guidance-adherence skill and follow its instructions to obtain a guidance adherence score for the theory.
4. Score Soundness: Invoke the score-soundness skill and follow its instructions to obtain a soundness score for the theory.
5. Final Output: Report the theory ID together with each score value (length score, guidance adherence score, and soundness score). Do not include any additional commentary in your output.