The Agent Skills Directory

[SAFE]: Secret Protection: The skill enforces a strict policy where the agent never handles literal API tokens. Credentials are managed through environment variables or user-created .env files, and their validity is checked indirectly via CLI exit codes. The skill also proactively verifies that .env files are gitignored before suggesting their use for token storage.
[SAFE]: Resource Integrity: All external tooling and delegated skills (e.g., @lingui/cli, glossary-generation) are official vendor resources or well-known community libraries. The skill uses a manifest-snapshot to freeze these dependencies at planning time to ensure a consistent and verified execution environment.
[SAFE]: Synchronization Security: Before creating new CI workflows, the skill checks for existing 'l10n*' branches using 'git ls-remote' to detect and avoid conflicts with pre-existing automation tools like the native Crowdin GitHub integration.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes external project source code to identify internationalization candidates.
Ingestion points: The skill reads the repository's source tree (e.g., src/ directory) to identify and count hardcoded strings in JSX, templates, and data modules.
Boundary markers: None explicitly present for user-supplied string content; the skill is designed to ingest and process arbitrary source text.
Capability inventory: The skill has the capability to write configuration files (.crowdin/, crowdin.yml, workflows), perform git commits, install NPM packages, and execute Crowdin CLI commands.
Sanitization: None; the skill preserves the content of the strings it identifies to ensure they can be accurately translated by the user or AI service.

i18n-setup