The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/generate_all_offline.py uses the exec() function to execute Python code read from the file step3_filled_placeholder.py. This allows the skill to dynamically load and execute logic at runtime, which is a significant security risk if the source file is modified by untrusted input.
[COMMAND_EXECUTION]: The SKILL.md file contains instructions for the agent to execute several shell commands. These include scripts from its own directory (scripts/generate_all_offline.py, scripts/merge.py) and scripts from other skill directories like paper-workflow-orchestrator, paper-formal-writer, and context-memory-keeper. This execution chain grants the skill control over various system processes.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources including tasks.json, writing_plan.json, and step3_filled_placeholder.py. These inputs are interpolated into templates in scripts/generate_all_offline.py and scripts/merge.py without sanitization or boundary markers. This creates a surface where malicious instructions embedded in the data could influence the behavior of the agent or downstream skills in the paper-writing workflow.
[EXTERNAL_DOWNLOADS]: The script scripts/merge.py contains a dependency on the python-docx library. If this library is not present, the script attempts to import it, representing an external dependency required for full functionality.

paper-micro-unit-generator