The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes Python scripts (checks.py and merge.py) for performing analysis. Technical review of these scripts shows they use standard libraries (html.parser, re, pathlib) to process text and do not perform any dangerous system calls, unauthorized file modifications, or arbitrary command execution.
[EXTERNAL_DOWNLOADS]: The documentation references research and guidelines from well-known technology organizations and government services (Google, GOV.UK, Baymard Institute, WHATWG, etc.) as the basis for its rubric. It recommends installing the jsonschema library, which is a standard, widely-used package for data validation.
[DATA_EXFILTRATION]: The skill does not perform network operations or access sensitive local files (such as SSH keys or environment variables). The analysis process is self-contained and operates only on the provided HTML artifacts.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests untrusted HTML data, its logic is restricted to heuristic analysis and producing structured JSON findings. It does not use the ingested data to influence its own instruction set or execute remote scripts, mitigating risk from malicious content within analyzed forms.

critique-forms