The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes and displays competitor data (including reviews, posts, and social media comments) scraped from untrusted external platforms. This content could contain malicious instructions intended to manipulate the agent's logic or behavior.
Ingestion points: External data is fetched from the Apify API via reference/scripts/run_actor.js and is either displayed in chat or saved to local results files.
Boundary markers: Content retrieved from external sources is not enclosed in specific boundary markers or accompanied by defensive instructions to ignore embedded command patterns.
Capability inventory: The skill workflow utilizes shell command execution (via mcpc and node) and file system access, which are high-leverage capabilities reachable from a compromised context.
Sanitization: The provided Node.js runner script includes security logic to prevent CSV cell injection (Escaped =+@- characters), but it does not sanitize retrieved natural language data for embedded prompt injection markers.
[COMMAND_EXECUTION]: The workflow requires the agent to execute shell commands, specifically leveraging the mcpc CLI tool and a local Node.js runner to manage API tokens and automate data extraction.
[EXTERNAL_DOWNLOADS]: The skill depends on the @apify/mcpc package and makes network requests to the apify.com domain (specifically api.apify.com and mcp.apify.com). These references are documented as interactions with a well-known service provider.

apify-competitor-intelligence