The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, including LinkedIn posts and community signals via common-room, which creates a surface for indirect prompt injection.
Ingestion points: Digital intent signals from common-room, prospect data from zoominfo and apollo, and content from a buyer's recent LinkedIn posts.
Boundary markers: The skill does not explicitly use delimiters or instruction-ignore warnings when interpolating external content into prompts.
Capability inventory: The skill invokes several tools (Epiphan AI, Brand MCP) and performs file write operations to the local filesystem for analytics.
Sanitization: While the skill uses check_my_copy for brand consistency, there is no evidence of security-specific sanitization or escaping of external data to prevent prompt injection.
[COMMAND_EXECUTION]: The skill orchestrates multiple external plugins and MCP tools, including common-room, zoominfo, apollo, and Epiphan AI, to perform research and draft generation.
[DATA_EXPOSURE]: The skill writes session outcomes and analytics data to absolute paths in the user's home directory (~/.claude/skill-analytics/ and ~/.claude/skill-runs/). While these are intended for internal tracking, automated writes to these locations should be monitored.

linkedin-social-selling-skill