The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions include a mandatory workflow section that directs the agent to spawn specific sub-agents and execute parallel orchestration tasks. This attempts to strictly define the agent's internal reasoning process and workflow flow. Evidence: The 'Agent Workflow (MANDATORY)' section in SKILL.md. \n- [INDIRECT_PROMPT_INJECTION]: The implementation patterns for locale detection ingest data from potentially untrusted external sources such as request headers and URL path segments. \n
Ingestion points: Accept-Language header processing and [locale] path segment extraction in proxy.ts and references/diy-locale-detection.md. \n
Boundary markers: Code examples include validation logic using routing.locales.includes(locale) to constrain input. \n
Capability inventory: The skill utilize file system operations (fs.readFileSync) and command execution (bunx next-intl extract) in utility scripts. \n
Sanitization: The patterns use explicit membership checks against a configuration-defined list of supported locales. \n- [DYNAMIC_EXECUTION]: The skill documentation encourages the use of dynamic import statements to load translation dictionaries based on runtime variables derived from user requests. Evidence: references/installation.md and references/configuration.md show await import(../../messages/${locale}.json). \n- [COMMAND_EXECUTION]: The documentation includes instructions for running external utilities using a package runner. Evidence: references/extraction.md contains the command bunx next-intl extract.

nextjs-i18n