The Agent Skills Directory

[Data Exposure & Exfiltration] (SAFE): The skill handles local file operations (reading and writing CSV/JSON files) consistent with its stated purpose. No network requests to external domains or access to sensitive system files (e.g., SSH keys, env files) were identified.
[Indirect Prompt Injection] (LOW): The skill ingests untrusted data via load_csv.
Ingestion points: load_csv method in StockScreener class.
Boundary markers: Not explicitly defined in the provided snippets.
Capability inventory: Local file read/write (CSV/JSON); no network or system command execution.
Sanitization: No explicit sanitization of CSV content is shown before data is returned to the agent context. While this is a common surface for indirect injection if the CSV contains malicious instructions, the tool's focus on numeric filtering makes this a low risk.
[External Downloads] (SAFE): Dependencies are restricted to reputable, versioned packages from the standard Python ecosystem (pandas, numpy). No remote scripts are downloaded or executed during runtime.
[Command Execution] (SAFE): The skill does not invoke subprocesses, shell commands, or dynamic code execution (eval/exec).

stock-screener