The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill fetches and processes data from various external platforms (Instagram, LinkedIn, TikTok, etc.) which is then returned to the model context. This content is untrusted and could contain hidden instructions.
Ingestion points: index.ts via ApifyDataset.listItems() and ApifyDataset.getAllItems().
Boundary markers: Absent; scraped content is interpolated directly into objects or strings returned to the agent.
Capability inventory: The skill can execute local scripts via bun and make network requests to the Apify API.
Sanitization: None detected; scraped data is returned in its raw form.
[COMMAND_EXECUTION]: Mandatory shell command execution. The SKILL.md and Workflows/Update.md files mandate the execution of a curl command to a local notification service (http://localhost:8888/notify) whenever the skill is invoked.
[PROMPT_INJECTION]: Instruction Override. The skill documentation uses high-priority markers ("🚨 MANDATORY", "REQUIRED BEFORE ANY ACTION") to force the agent into specific behavioral patterns, such as sending notifications before processing user requests.
[REMOTE_CODE_EXECUTION]: Dynamic Script Generation. The web-scraper.ts module allows the agent to generate and submit a JavaScript pageFunction for execution on the Apify platform. While this code runs in a remote sandbox, it represents a pattern of dynamic code generation and execution triggered by the agent.

Apify