The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a Python script, scripts/nda_clause_reviewer.py, which is used to analyze the text of NDA documents. This script is implemented using standard Python libraries (re, argparse, json) and operates purely as a text-processing utility. It does not perform any network operations, access sensitive system directories, or execute external system commands.
[PROMPT_INJECTION]: The skill demonstrates a surface for indirect prompt injection because it ingests untrusted text from external NDA documents via the scripts/nda_clause_reviewer.py script. Analysis shows that while ingestion points are present, there are no explicit boundary markers or sanitization routines implemented to separate the document content from the agent's logic. However, since the script possesses no dangerous capabilities—lacking the ability to perform network exfiltration or execute shell commands—the risk remains negligible and inherent to the skill's primary function of document review.
[SAFE]: Comprehensive analysis of the provided instructions, reference guides, and script files reveals no evidence of malicious behavior. The skill does not contain hardcoded credentials, obfuscated code, or unauthorized data access patterns. Its implementation is transparent and directly supports its stated goal of providing contract review assistance.

nda-review