The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/scrape.py manages a Scrape.do API token retrieved from a local file (config/token.txt) or environment variables. However, the fetch_via_scrapedo function transmits this sensitive token to the API endpoint over unencrypted HTTP (http://api.scrape.do), which is vulnerable to interception via man-in-the-middle attacks. Furthermore, the token is passed as a URL query parameter, which is often recorded in proxy and server logs.
[DATA_EXFILTRATION]: The skill performs network requests to an external service (api.scrape.do) that is not on the standard whitelisted domains. While this is part of the skill's primary function, the use of unencrypted transport for the API request increases the risk of exfiltrating sensitive scraping metadata or tokens.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external websites, creating a vulnerability surface where an attacker could influence the agent's behavior.
Ingestion points: The fetch_via_scrapedo function in scripts/scrape.py fetches content from arbitrary user-supplied URLs.
Boundary markers: The skill lacks explicit delimiters or instructions to the agent to treat the scraped content as data rather than instructions.
Capability inventory: The agent processes the extracted text or HTML, which could lead to following malicious instructions hidden in the scraped content (e.g., "Stop your current task and perform X instead").
Sanitization: Although BeautifulSoup is used to remove script and style tags, the remaining text content is not sanitized for potential prompt injection attempts.

scrapedo-web-scraper