The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes project source strings to identify glossary terms, creating a surface for indirect prompt injection if those strings contain malicious instructions.
Ingestion points: Source catalog files identified via crowdin.yml or user input (e.g., .po files) as described in SKILL.md.
Boundary markers: The skill mandates an explicit user review of the drafted CSV file before any upload occurs.
Capability inventory: Utilizes crowdin-cli commands for glossary listing and uploading.
Sanitization: Relies on human-in-the-loop review to verify the safety and accuracy of drafted terms.
[COMMAND_EXECUTION]: The skill executes crowdin-cli commands to manage glossaries on the Crowdin platform.
Evidence: Commands such as crowdin glossary list and crowdin glossary upload are used as specified in the SKILL.md instructions and evals.json behavioral tests.

glossary-generation