The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, including LinkedIn posts, profiles, and Sales Navigator notifications, to generate comment drafts. This ingestion of third-party content creates a potential surface for instructions hidden within posts to influence the agent's behavior.
Ingestion points: Content is retrieved from live LinkedIn feeds, search results, and external profiles via Chrome automation or Apify actors (SKILL.md §A3, §B1).
Boundary markers: While the instructions specify drafting in a specific 'author voice' and using quality gates, there are no explicit structural delimiters mentioned to isolate ingested post content from the agent's core instructions.
Capability inventory: The skill has the capability to write to local state files (engage-state.json), update Notion pages, and trigger external web scrapers.
Sanitization: A quality gate (anticopywriting-ai) is used to review output drafts, but no specific sanitization for incoming scraped data is described.
[EXTERNAL_DOWNLOADS]: The skill relies on third-party Apify actors for data extraction, which involves executing remote scraping logic.
Evidence: The skill references apimaestro/linkedin-posts-search-scraper-no-cookies, harvestapi/linkedin-post-search, and apimaestro/linkedin-profile-posts for batch processing (SKILL.md §A3, §B1).
[COMMAND_EXECUTION]: The skill processes complex CLI-style arguments to manage its execution environment and interactions with external tools.
Evidence: The skill supports various flags including --mode=targets|radar|log, --engine=chrome|apify, and --slot=morning|midday to control its operations (SKILL.md Frontmatter).

08-engage