The Agent Skills Directory

[PROMPT_INJECTION]: The skill operates on untrusted external content (articles from URLs or files) which serves as an indirect prompt injection surface. The risk is mitigated by the structured workflow and specific translation principles that constrain the agent's focus. Ingestion points: Processes user-provided URLs and files for translation as described in workflow-mechanics.md. Boundary markers: Employs Markdown formatting and a dedicated subagent prompt template (references/subagent-prompt-template.md) to maintain task focus. Capability inventory: Uses scripts/chunk.ts for file manipulation and triggers subagent spawning for parallel processing. Sanitization: The skill relies on model-level task constraints rather than explicit input sanitization.
[COMMAND_EXECUTION]: The skill invokes local TypeScript scripts (scripts/main.ts and scripts/chunk.ts) using the Bun runtime to perform markdown segmenting. These scripts use standard file system APIs to manage intermediate translation chunks and do not involve unauthorized system access.
[EXTERNAL_DOWNLOADS]: The skill utilizes the well-known markdown-it package as a dependency and suggests using standard runtime managers like npx to ensure the required execution environment is available. These are established developer tools and pose no inherent security risk in this context.

baoyu-translate