The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts scripts/analyze_page.py and scripts/competitor_scanner.py are designed to fetch and analyze content from arbitrary URLs provided as command arguments.
[COMMAND_EXECUTION]: The Python scripts scripts/analyze_page.py and scripts/competitor_scanner.py contain insecure SSL configurations. Specifically, they set ssl.verify_mode = ssl.CERT_NONE and ssl.check_hostname = False. This disables SSL/TLS certificate verification, allowing the agent to connect to servers with invalid or malicious certificates. This exposes the user to Man-in-the-Middle (MitM) attacks where an attacker could intercept data or inject malicious content into the analysis flow.
[PROMPT_INJECTION]: The skill architecture is susceptible to indirect prompt injection through external data ingestion.
Ingestion points: Untrusted website data is fetched into the agent's context from scripts/analyze_page.py and scripts/competitor_scanner.py.
Boundary markers: The skill lacks explicit delimiters or instructions for the AI to ignore instructions embedded within the scraped HTML content.
Capability inventory: The agent has permissions to write files to the local file system (e.g., reports, audits) and execute Python scripts to generate PDF documents.
Sanitization: No sanitization is applied to the retrieved text or HTML attributes before they are processed by the agent, allowing potential attacker-controlled instructions on a webpage to influence the agent's behavior.

market