The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines a customMapFunction parameter which is described as a "JavaScript function to transform each output object." This creates a dynamic execution sink where user or agent-provided code is executed at runtime. Additionally, the skill instructions include the execution of local scripts and system commands.
Evidence: node scripts/run_actor.js is called multiple times to run the scraping logic.
Evidence: curl is used to interact with the Apify REST API, including passing the APIFY_TOKEN in the URL query string.
[PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection because it retrieves and processes untrusted third-party content.
Ingestion points: Raw tweet text and metadata are fetched from Twitter using the apidojo/tweet-scraper actor and presented to the agent (SKILL.md).
Boundary markers: The skill uses Markdown tables for output formatting but lacks explicit instructions or delimiters to warn the agent against executing commands embedded within the retrieved tweets.
Capability inventory: The execution environment allows for shell command execution (node, curl) and network requests to external domains (api.apify.com).
Sanitization: No sanitization or filtering logic is provided to handle potentially malicious instructions within the scraped tweet data.

scraping-tweets-by-keyword