The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines an attack surface by ingesting external UI strings from project files. It includes mandatory safety instructions for the agent to treat this content as data rather than instructions. Evidence: Ingestion points (locales/en.json), capability inventory (local auditing script), and explicit boundary markers in the 'Security boundaries' section of SKILL.md.
[COMMAND_EXECUTION]: The skill executes a local Python script (scripts/capcheck.py) using the terminal. The script uses secure parsing methods like yaml.safe_load() and performs sanitization of text via the clean() function. No suspicious network activity or privilege escalation patterns were identified.

ui-capitalization