The Agent Skills Directory

[SAFE]: The skill provides extensive data sanitization routines in references/data-transforms.md. These include mechanisms to remove zero-width characters (e.g., U+200B) and perform Unicode NFKC normalization, which are effective defenses against certain types of prompt injection and obfuscation attacks in extracted data.
[COMMAND_EXECUTION]: In SKILL.md, 'Strategy C' specifies the use of Bash and curl for fetching data from APIs and downloading files. This is consistent with the skill's primary purpose of web data extraction.
[DATA_EXFILTRATION]: While the skill is designed to extract data from external websites, the workflow is transparent and focused on structured data delivery to the user, with no evidence of unauthorized background data transmission.
[SAFE]: Regarding potential indirect prompt injection, the skill includes a complete chain of mitigation: it defines clear ingestion points (Phase 2/4 in SKILL.md), output boundary markers (delivery envelope template in references/output-templates.md), and robust sanitization (data cleaning transforms in references/data-transforms.md).

web-scraper