The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external company websites and internal CRM records to generate outbound communication, creating a surface for potential indirect prompt injection.
Ingestion points: Company research (Step 1b) and CRM search (Step 1a).
Boundary markers: The skill employs an evidence grading methodology (e.g., [CRM], [Verified], [Inferred]) to label data sources, which provides logical structure but lacks strict technical delimiters (like XML tags) for untrusted content.
Capability inventory: The skill possesses the ability to read local filesystem profiles, search CRM databases, and generate text-based and script-based artifacts.
Sanitization: There are no explicit instructions to sanitize, escape, or filter content retrieved from web research before it is interpolated into email sequences.
[DYNAMIC_EXECUTION]: The skill references the generation of PDF artifacts using Python and the reportlab library, which involves generating and executing code at runtime.
Evidence: Section 8 (Artifact Generation) describes Option C, which utilizes Python and reportlab to produce a single-page cheatsheet.
Pattern: This involves script generation from research templates, which is a standard feature of data analysis workflows but represents a dynamic execution vector.

gtm-account-snapshot