The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized data access were detected. The skill relies on pre-defined tool names and a controlled gateway for API interaction.
[INDIRECT_PROMPT_INJECTION]: The skill processes company data from external sources, which constitutes a potential injection surface. However, the risk is mitigated as the skill defines a read-only research scope and lacks high-privilege capabilities such as file-system writes or arbitrary command execution.
Ingestion points: External firmographic and regulatory data retrieved via tools in SKILL.md.
Boundary markers: Absent, but instructions emphasize citing sources and separating observations from analysis.
Capability inventory: The skill possesses no scripts or tools for file writing, network exfiltration, or command execution.
Sanitization: The workflow requires calling sandbase_describe_tool to validate parameter schemas before every tool execution, providing a level of structural validation.

company-enrichment