The Agent Skills Directory

[Indirect Prompt Injection] (MEDIUM): The skill's primary function is to ingest data from untrusted external sources (crawled websites) and use that data to generate files.
Ingestion points: scripts/main.py fetches external HTML via requests.get(url) and parses it with BeautifulSoup.
Boundary markers: None present. The script treats all discovered links as valid data for sitemap generation.
Capability inventory: The script performs network operations (requests.get) and writes files to the local disk (tree.write).
Sanitization: Limited to basic domain matching and extension filtering (e.g., .pdf, .jpg). It does not sanitize the content of the URLs or the response against malicious injection patterns.
[Data Exposure & Exfiltration] (LOW): The script performs network requests to arbitrary domains provided by the user.
Evidence: requests.get(url, headers=HEADERS, timeout=10) in scripts/main.py allows the agent to communicate with any external server.
[Unverifiable Dependencies] (LOW): The skill relies on several common third-party libraries.
Evidence: requests, beautifulsoup4, and lxml are standard but represent an external dependency chain that must be managed.

sitemap-generator