The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the project's codebase, including locale definition files and source code templates. This creates a surface where malicious content within these files could potentially influence agent behavior during report generation or file-scaffolding processes.\n
Ingestion points: Reads from locale files (e.g., messages/*.json, config/locales/*.yml, .po catalogs) and source code (e.g., .tsx, .vue, .html, .erb) as detailed in SKILL.md and references/frameworks.md.\n
Boundary markers: The skill does not define explicit delimiters or instructions to the agent to treat codebase content as untrusted data in its reports.\n
Capability inventory: The skill utilizes Bash for command execution, Edit and Write for file modifications, and Task for sub-agent management.\n
Sanitization: Data from processed files is extracted and reported verbatim without mentioned escaping or filtering mechanisms.\n- [DYNAMIC_EXECUTION]: The skill uses the shell to execute dynamically generated Python and Ruby code to perform data processing tasks.\n
The references/frameworks.md file contains logic for executing Python heredocs and Ruby one-liners via the Bash tool to flatten and diff locale file structures.

i18n-check