The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and processing account lists from external sources like CSV files, Google Sheets, and Notion databases. This ingestion point constitutes an attack surface for indirect prompt injection, as malicious data within these lists could attempt to influence the agent's research agents or output formatting. This risk is inherent to the skill's purpose of data enrichment and is mitigated by the structured processing phases described in the instructions.
[EXTERNAL_DOWNLOADS]: The instructions direct the agent to interact with a wide array of external B2B data providers and public datasets (e.g., LinkedIn, Apollo, Crunchbase, Dodge, and municipal permit portals) to perform account enrichment. These references target well-known industry services and official public data sources, representing standard business intelligence operations.
[COMMAND_EXECUTION]: The skill implements dynamic directory resolution by globbing for its own metadata file (SKILL.md) to establish its base path. This practice avoids hardcoded file paths and ensures that resource loading is handled safely and reliably across different execution environments.

signal-research