The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data from LinkedIn posts and comments. While this presents an attack surface where third-party content could contain malicious instructions, the skill implements robust defensive measures.
Ingestion points: Untrusted data enters the agent context through lib.fetch_post, fetch_post_comments, fetch_user_recent_comments, and fetch_post_engagers (SKILL.md).
Boundary markers: The skill contains a comprehensive 'Untrusted content' section that explicitly instructs the agent to treat all fetched text as data rather than instructions, providing clear rules for handling such content.
Capability inventory: The skill uses the ingested data to generate structured reports and draft outreach messages. It does not perform high-risk operations like arbitrary code execution or sensitive file modification based on this data.
Sanitization: The skill employs natural language instructions and references a dedicated security guideline (../../references/untrusted-content.md) to ensure the agent maintains safety boundaries.
[EXTERNAL_DOWNLOADS]: The skill interfaces with Apify, a well-known web scraping and automation platform, to retrieve social media engagement data.
Source: Data is fetched via the Apify MCP connector or lib.ApifyClient.
Context: The use of a specialized third-party service for data extraction is a core functional requirement and avoids the need for the agent to manage raw LinkedIn authentication sessions.

linkedin-engager-analytics