The Agent Skills Directory

[SAFE]: No security issues or malicious patterns were detected. The skill's operations are limited to local file analysis and providing writing recommendations.
[COMMAND_EXECUTION]: The skill executes a bundled Python script to calculate document statistics.
Evidence: scripts/section_budget.py is called within SKILL.md to map length.
The script is deterministic and uses only Python standard libraries (argparse, json, re, sys).
No dangerous functions such as eval(), exec(), or subprocess.run() are present in the script.
[DATA_EXFILTRATION]: No network access or data exfiltration patterns were found.
The script reads user-provided .tex files locally and outputs text/JSON to the console.
Persistence is limited to local project directories (.paper-memory/ and paper-workspace/writing/) for session logging.
[PROMPT_INJECTION]: The skill processes untrusted LaTeX source files for analysis.
Ingestion points: scripts/section_budget.py reads the content of the texfile argument.
Boundary markers: The script strips LaTeX comments and markup before analysis, reducing the risk of accidental instruction following.
Capability inventory: The skill can recommend edits and write to the local filesystem, but provides guardrails stating the author must approve all changes.

fit-page-limit