The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted text from patent claims and abstracts retrieved via MCP tools into the patent_markdowns/ directory. There are no explicit boundary markers or sanitization steps mentioned to isolate this data from the analysis prompt. However, the agent's capabilities are limited to analytical processing and local file/directory management, and the skill does not instruct the agent to execute commands based on the retrieved content.
[DATA_EXPOSURE_EXFILTRATION]: The included Python script scripts/create_landscape_project.py implements a security check using the ensure_within function to prevent path traversal attacks, ensuring project directories are created only within the designated root folder.
[SAFE]: No malicious patterns, such as command injection, hardcoded credentials, or unauthorized network activity, were found. The skill follows best practices for tool usage and project organization.

map-small-rna-patent-landscape-ls