The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to run a Python script (scripts/search.py) using the agent's Bash tool to perform Google searches and extract web content.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the trafilatura library and makes network requests to the Serper API (google.serper.dev) and various external websites discovered during the search process.
[PROMPT_INJECTION]: The skill extracts full-page text from third-party websites, creating a surface for indirect prompt injection. Malicious instructions embedded in the processed web pages could potentially influence the agent's subsequent reasoning or actions.
Ingestion points: The _extract_content function in scripts/search.py fetches and processes data from external URLs.
Boundary markers: The extracted content is returned within a structured JSON object, but the text inside the content field is raw and unsanitized.
Capability inventory: The skill can execute Python scripts and perform network operations via the Bash tool.
Sanitization: There is no evidence of sanitization or filtering of the extracted text to prevent instructions from being interpreted by the agent.
[DATA_EXFILTRATION]: The script scripts/search.py contains a hardcoded absolute path (/Users/kirk/Projects/openclaw-skills/.env) used to load environment variables. This exposes information about the author's local directory structure and could lead to unintended file access if a similar path exists on the system where the skill is deployed.

openclaw-serper