The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a bundled local script scripts/scorecard.py using python3 to compute paper scores deterministically based on extracted features. The script is safe, uses only standard libraries (json, sys, argparse), and performs no network or sensitive file operations.
[DATA_EXPOSURE]: The skill reads and writes to a local .paper-memory/ directory to maintain a profile of the paper and log improvement suggestions across sessions. This data is stored locally within the user's project environment and is not exfiltrated.
[SAFE]: The skill follows best practices for document-processing AI tools, including deterministic scoring logic separated from the LLM, transparent disclosure of corpus sources, and clear disclaimers regarding the scope of its analysis.

benchmark-paper